[FilSciHub Research University] Course #5: Basic Statistical Design of Experiments

okay bye bye [Music] [Music] do [Music] [Music] [Music] [Music] i think we can start marty [Music] why do you think we can start okay this saturday for another webinar for that is part of our research university component of the filipino science hub so kuya jp if you can share this type uh doc jeff are you there yeah yeah [Music] so for today we will be having our basic statistical design of experiments and this will be delivered by professor lara paul ebal of the institute of statistics of up los banos next slide please so before we start the webinar let me remind our participants to please mute their devices and also turn their video off of course unless you want to post a question at the end of the webinar you can press the racer hand button and you will be acknowledged by the moderator and uh we also request you not to record this session because all of these sessions will be uh uploaded in our fa on our youtube channel next slide please all right so for the certificate certificate of participation we will be issuing one uh after the webinar we will post a google form link to issue a certificate of participation for this session and this will be posted in our zoom chat box and youtube live comments and it will also be posted in the video description in youtube in case you missed the live session uh but uh certificates will be only issued to subscribers of our youtube channel and facebook page next slide please so i turned the floor to huya okay so please let me introduce the platform so filipino science sub is an online platform that we founded back in 2012 and its main intention is to promote the cultures of stem education and research among filipinos so at the moment oh this group is uh being led by a team of filipino scientists and educators um a graduate from the of uh chemistry from the institute of chemistry um at the at the up in spanish and then we all sought higher education for in different areas of stem and so we are all volunteers um dedicating our time to actually build uh a new culture a new generation of of scientists and filipino science enthusiasts um for the future so um what is it that we're trying to do so the filipino science hub intends to again build to create a new uh or to improve the culture of uh stem education and um research in the philippines so and and to us what what would make what would comprise a culture so the first of which is um we have to improve the mastery of our courses amongst filipino students and teachers [Music] you don't stop there uh for you to be able to launch a culture what the mind knows the hands should actualize so determine actualization and so these two components are actually delivered by two of our main programs so the first of which is filipino science hub education and filipino science uh um research university so uh in the succeeding slides we'll talk about this more in like in greater detail so we've been doing web events like this for the past 11 months or almost 12 months already and um so far but we have carried out about 26 events um so far and these events are comprised of teaching webinars uh career guidance talks web lectures research training courses like this one so this is i think the sixth one that we are rolling out and then um about five stem special topics um so so far paul we have our events um program has already rendered about 63 uh total training hours so 63 hours point of expertise speakers um and then uh we have already um educated about 14 000 live participants and today but we're anticipating like a fairly huge um turnout because this is a very relevant topic topic to both the high school and college um demographic and so far but we have rendered about 35 000 um training hours so um i'm going [Music] um okay so for today um i'm gonna hand the floor over to martin mateo because he's gonna talk about phil's i have education yeah thank you jeff so i'm back to introduce that education cluster of uh the filipino science hub so our main vision is to produce around or to impact around 27 million students in the future this is by targeting academic frontliners which is the main target of filipino science hub or young followers mostly are academic frontliners and they are in their early careers so if we project around 30 students for educator for their whole career of around 30 years we project to impact around 27 million students that can be stem practitioners and nation builders in the future next slide please so we have also launched teaching webinars this is uh in the onset of the pandemic so during those uh early days of the pandemic we launched uh learning through uh through distance learning or online learning so all of these resources are uploaded in our youtube channel so if you are a fairly new teacher on the field and uh you want to improve your uh skills on online teaching you can always visit our webinars in our youtube channel next slide please we also uh launched our modules so we have bio biology chemistry and physics mode modules curated by bs math and science teaching graduates and students from uplb so ita manga mojo warrior sam in a sobrang massive c i n g c i v and c mark those teaching guides can aid those modules can aid teachers and students alike so it consists of teaching guides class activities and problem sets and these are all accessible through our facebook page and our website that's www www.philsaihub.com so curated push upper subject madeline website next slide please all right so we will be launching a stem teaching fellowship where filipino stem educators get trained by pinoy scientists so this is a fairly new uh campaign of the filipino science sub to bring in more teachers to our team and to be uh trained by mentors from the stem that that are stem practitioners so what are the perks of being a phils i have ed fellow you will have a formal phils i have ed membership and member mentorship in practical science teaching so uh this is more of in experiential uh science teaching and you can also present your stem teaching materials on an online stem education platform and we will be uh helping you to improve your delivery of the subject matter especially on the part of the uh especially on the part of the content and the quality of your teaching materials and uh details on how to apply to fields i have ed fellowship program will be released on april 20 2021 and puya jeff will add something yes okay so um mainly because there's a couple of very important aspects to this so we are collaborating with the major for education foundation in the philippines and a major teaching university focuses on a leading um universities when it comes to teaching um and pedagogy and uh we're trying to come up with the mechanism in such a way that you'll be trained in terms of um basics of uh teaching stem and then courses for that we ourselves at filipino science will come up with paraphrase practical and experience stem courses so we're looking at about five to ten of course training courses and then after puno again practically quote unquote um internship where you will be asked to synthesize something out of the training materials that we provide you with and then at the end of this um there's a possibility for that we might be able to provide you with cpd units which i think is very relevant these days and all these um we are actually going to be offering for free design internships will actually you can do it at your own pace so if it takes you 10 months to finish this that's okay if it if you can actually finish it quite quick quite quickly then you can also do that and what's most important is that you learn you earn cpd units and all of these will be for free and then you will be um presenting your work in front of an esteemed panel of educators and scientists filipino filipino science education program and um um that we actually carry out or two weeks that we conduct at filipino science hub research university all right jp all right thank you doc jeff uh good morning everyone my name is jp onya and i am the head of the phil sci hub research university so uh as dr jeffrey mentioned affiliate research university is one of our core campaigns here at phil's ihub and in this program we are focusing on scientific research and its different aspects so tini train punatindito both the teachers and students uh the the entire range of a research research endeavor okay so here at fields i have research university we will share um some of the best practices in terms of uh conducting a research project and we will share our our own uh research experiences in doing so and also some tips and advices okay so philsaya research university is composed of eight mandatory courses in fact uh our webinar today is one of the mandatory course for research university which is statistical design our fifth so we have already completed uh four uh mandatory courses uh from research ideation literature review experimental design and uh research proposals so little until uh we we could complete the the entire uh course and we will commence with the conference and poster uh presentation webinar okay next slide so uh mandatory courses we will issue a certificate of completion and young mechanics certificate of completion is summarized in our online event last january to to launch the philharm research university so kindly uh check it out i will also post the link uh on the uh confirmation email uh the email issuing your your certificates for of participation for this webinar so kindly check it out and we also devised a google classroom for this and the link for that is also provided in that email so and the fields i have platform is always present in all of our channels mainly the the our website www.philslihab.com and we're also present on facebook youtube and even on tech talk and there in our main platform which is our website www.philsayhub.com it contains all of our webinars tutorials modules virtual lab and features so mamayapo [Music] and there we will share um like a virtual virtual what we did was we compiled um a collection of stem modules for in biology chemistry and physics so these are on in pdf formats on how to access those things is fairly straightforward for our teachers and students to access them and download them so watch out for that okay all right and i think on that queue marty you can take it okay thank you huya jp and kuya jeff so before i introduce i before we introduce our speaker to you let me give let me just get let me acknowledge our viewers so napoleon around 400 viewers a youtube and 100 here's a zoom so around 500 uh participants already and uh still growing uh to give you to introduce to you our speaker for today i get i give the floor to dr din diu thank you good morning salahat it's my honor to introduce our speaker for this morning professor lara embal she is currently a assistant professor at the institute of statistics at ub los panos and um she is uh also um she's she she got her bachelor's degrees uh degree instead so this is from indiana state university university in illegal city and uh she got her master of science in statistics at uplb and i believe she's currently finishing up her phd in stat in the same university with a focus on cognitive business management she is active an active member of instant instead statistical consulting group philippine statistical association and international statistical institute so let's all welcome professor laura eppa mom take it away good morning paul good morning for everyone uh thank you for having me here as a resource person of this webinar lecture series um it's my honor to be part of this lecture series and will i share now my screen yes ma'am please okay so yes screen okay so before i formally start may i know if you can hear me yes ma'am clearly can you see me yes can you see my slides yep yep yeah okay thank you so thank you for letting me know so before i formally start this um uh lecture okay so i would like to thank um the organizers of this webinar lecture for having me as a resource person and then i would like to thank insta for allowing me to share my knowledge on this experimental design and i would like to thank everyone for giving time okay to attend to this webinar on basic statistical design of experiments so now um before again to formally start let me ask you this question what comes to your mind when you hear the phrase is that experimental design so you may okay participate okay you may go to mentee.com and enter this code okay so i'll stop share first can you see the code nine six two nine two eight five three okay so nine six two nine two eight five three okay so i will just wait i will screen share my mentee okay okay so nine six two nine two eight five three so i will wait at least now i i need to know the level of your knowledge on this experimental design so that i'll be able to discuss you know in a way that you want it to be okay so research statistics data process drawing okay hello okay thank you for participating so i see many responses here so control group ib dv so iv is independent variable dv is dependent variable statistics okay common data statistics research setups trials methods randomize placebo group layout trials grouping randomization lots of math grabbing among lots of math variants that's very important framework okay again that's actually the layout okay crd yeah i sense that some of you here have a good background i think in research or experimental design or in statistics some are just starting to learn you know statistics or experimental design or some are just starting uh doing research i think okay [Music] so thank you for letting me know so far we have 75 76 participants in this activity okay thank you for your time no effort menti.com pero thank you very much at least i have the idea so my anova yes that's very common um rcbd yeah i see rcbd crd okay uh variable process difficult oh that's sad okay but it's okay i will try to simplify things so that you'll be able to appreciate no experimental design um hypothesis i see hypothesis pattern yeah quantitative yes that's true quantitative so we're dealing with data and quantitative um experiment treatments okay so jung statistics and research in a way no that comes to your mind the uh when you think of experimental design okay so we've reached uh 102 participants so participate hopefully no pero although you have three uh chances there not to answer this so okay so will i stop here now so i think this is a good anona at least this gives me an idea of okay of your level of knowledge in experimental design so i'll stop sharing now thank you for your participation um it's okay so okay so maybe i'll wait sir jp um i think it's okay but i will just uh mention uh to everyone questions uh you can post them in the chat box here in zoom and also on youtube and then we will collect them at uh paraposa discussion later so at any point paul during the the lecture you can you're free to post any question or comment and we will gather them thank you sir jp so i'll stop here in a lang and then you can just ask sir jp or in the chat box yes okay okay i'll stop now and then i'll go back to my slides okay so at least now i know your level of knowledge in experimental uh design so i have uh okay this outline here so that you'll be able to at least know the flow of my discussion so number one i'll start with the basic concepts so basic concepts um includes include this uh basic terms that we need to learn okay and then we also have these principles and then we also have this um basic design structures that we need to learn so this actually include the crd the rcbd and the lsd so these are just basic uh design structures but they are very important in our experiment or research okay so we'll have a break also after uh before we discuss the rd here in the next topic before we discuss crd uh we'll have a break and then after we discuss the c rcbd we'll have a break okay so we have to break so don't worry so just imagine that this will take like uh more than an hour okay i hope no um no so we'll see okay so i will start now so i'll start with this uh basic definition on research or specifically scientific research so scientific research is a study conducted in contribution to science by the systematic collection interpretation and evaluation of data so we need to have the systematic collection why because we need to come up with a valid data for us to have a valid interpretation and valid analysis okay so that's why we must have to be we must be systematically uh we must systematically plan okay before the conduct of the experiment or the research okay so you are asked actually by your advisor or your teacher or your science teacher that you have to consult not to a statistician okay i hope you agree with me right so do you hear that some other teachers or other advisors tell you uh tell you the student to okay talk to your statistician or discuss this with your statistician so then the question is really what are the roles of statistics in a scientific research so we have this um roles number one we have uh designing we have designing experiments and surveys and then collecting data or summarizing the data and then describe data estimate the parameters and that test the hypothesis so very basic on designing of experiments uh before the collection of the data because again we need to have a valid data so to come up with a valid data we need to plan properly or experiment or we have this designing of experiments okay so before we discuss further this designing experiment we need to come up we need to understand some terms so the common terms that we usually encounter in doing research okay is we have this treatment or treatments experimental unit or sampling unit and then the response variable so maybe some of you are already familiar with these terms but maybe those who are just starting not their research or learning statistics so um we need these terms now we need to understand these things okay start moving okay so just to give you a picture of the relationship of those terms okay so we have this treatment so how is this treatment related to experimental unit so you can see the picture there so you can see that the treatment is applied to the experimental unit okay and then later on we gather the data so that's your response or responses okay so this is how we picture out and then okay so to formally define a treatment okay so treatment or factor sometimes you call it a factor includes the different procedures or conditions to be compared okay so you have the intention to compare those procedures or conditions okay or formulations or applications or methods okay and these procedures or conditions are actually your treatment levels okay so these are the preset quantities of a quantitative treatment or categories of a treatment under study okay so example okay so just to give you an idea so we have this type of sugar so you have you have the intention or you really want to compare this uh different types of sugar in terms of sweetness now when they are applied to cakes okay so this types of sugar which is uh which includes the icing the granulated and the dark muscovado so these are the types that you want to compare so these are your treatment levels another example is a type of irrigation so this is very agricultural example um you want to compare the different types the surface the sprinkler the drip and the subsurface okay so how do you compare or how do you compare this type so you need to apply this to for example rice okay and then later on we get the yield okay so we can now compare in terms of yield okay another example is the teaching method so this is very common i think nowadays okay you want to compare the performance of your students not say math score so you compare the teaching method online face to face and mix online in terms of their performance okay so just to give you an idea i know levels so depending on study search so these are just my examples another example another term that we need to learn is this experimental unit so the eu so unit or group of units to which treatment is applied once okay so you applied you know the treatment okay you know your treatment level you apply it to a subject or to a unit so we call that experimental unit so once kalama treatment experimental unit and then example we have blood of land a class of students or a plant or a tree or a dish no or a flask okay in the laboratory or a test okay and then another term is the sampling unit so it's a portion of the experimental unit on which the response variable is observed and measured so meaning to say it's just a portion so you cannot afford to measure the whole experimental unit or the whole plot so you just need to get some subsamples okay okay so we call them sampling units so example a seedling in a seed bed a tree in a group of trees a sample in a batch okay take note here wait i will move this sampling unit may be the whole experimental unit sometimes our research is up to the level of experimental units so we don't do sub samples or we don't get sub samples okay so that's it so sometimes same like your experimental unit with that sampling unit okay another term is the response variable okay so basically these are just the outcomes that you observe no in the field or in the laboratory so example yield of the crop height of the plant gain and weight volume of sales or math score or any other variables or any other data okay okay so we have this a simple example here just to the test know how you understand these three uh these terms okay so determine the treatment the response variable the experimental unit and the sampling unit okay number one an experiment was conducted on the effects of three sulfur applications so we have three levels 300 600 1200 on treating scab disease in potatoes so each application were randomly assigned to five plots so if you can imagine we have one application applied to how many plots five okay and this percentage of surface area of potato that was infected with scab or we call it scab index was measured from each blood okay so based on this example can you determine the treatment what's the treatment of the study for applications and then response variables so what's your data here okay your scab index and the next is the experimental unit so where do you apply where did you apply the the applications i sold for applications so to that five plus or to the plots so that's your experimental unit now okay how about something unit do you have something unit here okay so let's see we have i know something you need in that experiment or example so just to summarize again the treatment in the example is the surface application and then the treatment level so we have this specific levels 300 600 and 1200 and then the response variable is your scalp index or the percentage and then the experimental error is where you applied your treatment so you know plots and then sampling unit is the or none or there's no sampling unit okay another example a researcher wanted to determine if there were differences in the amount of calories among the three types of hot dogs so for example we have these three flavors beef pork and chicken so he selected three hot dog packages from each type and took three pieces of hot dogs from each package in which he took measurements of the amount of calories so i know so what happens now is that we have these different types and then we have from each type we have different packages however we're not able to to measure we cannot measure all the packages okay so what we get what we do is we get only few or say three hot dogs okay so now determine the treatment the response variable experimental unit and the sampling unit is there something you need here okay so think or read again the problem okay so i'll show you know the answer okay so you can see the problem here in the in the left hand side okay so the answer okay we have treatment the types of hot dogs and then the treatment levels we have the specific um flavors or pork chicken beef and then the response variable is the amount of calories and then the experimental unit okay that's the packages the hot dog packages okay and then now here comes the sampling unit which is or which are just the three pieces of hot dogs from each package okay so you can now really know my wedding uh there's an experiment that up to the level of experimental unit then there is also another experiment which is up to the level of a sampling unit okay so another term that we need to learn is experimental error so this is the variation in the observed values of the response variable from the experimentally illustrated align so if you can imagine there is this treatment level or condition or procedure applied to an experimental unit okay but take note you do not only apply uh once okay the the treatment level to only one experimental unit you have to consider uh more experimental units okay and then you also expect that these experimental units must be uh have the same responses or outcomes or at least they are close to each other right okay so now if they are different or there is this discrepancies or differences among these experiments or responses from these experimental units then that's incorporated in the experimental error okay so we need to have a measure of experimental error that's why we need to come up with many um experimental units or replicates okay okay why even if we try to to control and even if we try to only apply one kind of treatment level in several on several experimental units okay they still differ because of okay some reasons number one inherent variability of the experimental materials maybe because of the material used no um errors on in experimentation so it is usually the human error okay it's the person who is measuring the or gathering the data then errors in observations or in the measurements okay so these are some sources of um discrepancies no then we also have this combined effects of all extraneous factors so there are other factors that we need to to know at least we need to have a good background of that certain kind of research at least not into iman factors that can contribute to that response okay so we need to also try to control them okay because they actually uh the source of experimental error okay so just to illustrate okay so i think you're already familiar with this um picture or diagram okay so the experimental comes here okay so this values here the discrepancies between y 1 1 the y 1 2 and the y 1 3 okay so the discrepancies again are incorporated in the experimental error now another example okay so under experimental errors i'll just show you um an experiment was conducted it's the same example the potato example that i gave you okay the experimental error there consists of the pulled variation okay the pull discrepancies now among the five plots so what are these five blocks you know experimental units nothing in the same sulfur application so in the same level of sulfur say 300 or 600 or 1200 so at least now now we learned that we have this experimental error defined in this way so it's the full variation okay another term that we need to learn is sampling error so if we have experimental unit okay experimental okay we have this experimental error then if we recall we have the sampling unit so we also have the sampling error okay so sampling error is a measure of variation among sampling units within an experimental unit so if you can imagine we conduct or we conduct sub-sampling or we use sub-samples okay when sometimes we cannot measure the whole plot or the whole experimental unit so sometimes we get three or more okay so those variations of the responses of the sampling units okay are incorporated in a man in the sampling error okay so that's it so we have experimental error we also have something error now we have an example here just to give you an idea okay so recall we have this hot dog example we have three types of hot dogs and then we get three packages from each flavor and then from each package we get how many three hot dogs okay so now the variation of these three hot dogs the differences in terms of calories okay are incorporated now in the sampling error so sampling error consists of the full variation among the three hot dogs taken from the same package okay that's it now another term is precision so i think we're all familiar with this but we'll just have a quick review on this now um it's the ability of a measurement to be consistently reproduced okay as you can imagine again we apply the treatment to the experimental units to several experimental units and then these measures are coming from this measurements coming from this experimental units must be close to each other so that would mean we have we are getting precise estimate okay okay so again precision is measured by variance so we have to to compute for the variance in order for us to know if we have a high precision or low precision so ways to increase precision one to increase the number of samples this is very common way okay it means that you can increase your experimental units or you can increase your replicates another is skillful grouping of experimental materials so the researcher should have a way you know should have this skill no you should have a good background of your research okay to really know i know treatment okay that will also increase the precision and then aside from position you have accuracies it's the closeness of the observed values to the true value okay so if precision is the closeness of this uh measured values accuracy number is the closeness of that value to the true value so sometimes this true value is the true mean okay so this is measured by bias okay and then ways to increase accuracy okay we refine the experimental technique and we have to properly select the treatments okay now another term is the layout this refers to the final arrangement of the treatments okay over the whole experimental area so as you can see in the right hand side we have this example okay of this arrangement of treatment so for example b1 or the block one we arrange the treatments as a b c d and then for the next block we have this arrangement d a b c and then the third block c d a b and then the fourth block b c d a so this is an example of a layout okay so another term maybe this is new to you or i don't know um we should understand this too because this is very helpful for helpful in your experiment so fixed effects all factors under the test are fixed factors so a nuance a fix a factor is considered fixed when its levels are selected on purpose so you really have the intention to compare these factors example you have t1 t2 t3 and then the interest is on the effects of this selected level of fixed factor no inferences shall be directed at any treatment level that is not included in the experiment okay this means that when you have your conclusion your conclusion will only be applied to the other say will only be applicable to other experiment or research using the same set of levels so t1 t2 and d3 so that's under fixed effects now for random effects should break women fix my random so all factors under test are random factors a factor is considered random when the levels of the factors tested are a random sample from the population of levels so example as you can see here in the picture you have many levels so we have t1 t2 t3 t4 t5 and d6 then we just randomly select okay whatever intention my specific uh treatment level i was to may compare you just randomly select t4 t3 and t6 so it's fine okay so what you are you are doing is actually run under random effects so interest is on the population of treatment levels from which the random sample of treatment levels was drawn okay as you can see in the picture then population treatments will then be described in terms of the sample treatments okay so just to give you an idea of this too we have this example a drug study used a 5 0 mg 5 mg and 10 mg of an experimental drug okay say take note my zero my five by ten so under fixed effects the conclusion is applicable to other studies or treatments that use the same values of drug okay so applicable in conclusion more nano gamma zero five and ten okay so those other studies which also use this zero five and ten okay if you have that intention to compare zero five and ten now for random effects the conclusion is applicable to other studies that use other values beyond this particular values so um this 0 5 and 10 were just randomly selected okay and then yo so whatever the conclusion from that study okay will also be applicable to other studies uh not even other studies which actually do not use this zero five and ten so quite the only randomly selected zero five and ten okay so these are the terms that you learn i hope no so treatments experimental units sampling units response variable experimental error sampling error precision accuracy fixed effects and random effects okay so we'll have a okay next set okay of topic so we go back to this in statistics so i hope you'll understand those you have unders to those uh terms okay um now we go back to this in statistics a procedure of a randomized experiment is governed by the experimental design so um experimental design but okay we have this term randomized experiment okay what is this experiment proactively it's a randomized experiment okay let's see what is a randomized experiment so take note first if we have experiment it is characterized by the following number one presence of treatments okay you know already what is treatment experimental units okay and by the way the treatments are assigned to the units okay and by the responses that are measured so you have these terms already now you are all familiar with this um the next is the randomized experiment so it's an experiment where assignment of units to treatment groups are randomized so there is this randomization occurring in this kind of experiment so goal is to see the effect of the treatment while controlling for other factors so you have you should have a good backbone again of your experiment so that you can control them those other factors that kind of that can affect your outcome or response so that you'll you can focus on the comparison of the effects of the treatments that you really want to compare okay i hope you get me know so you have to control so that you can focus again to other treatment levels that you want to compare now present the strongest support to causation so this randomized experiment presents stronger support this means to say that this randomized experiment ensure ensures us that the treatment is really have an effect on the outcome or the response okay that's why we have to yun we have to conduct this randomized experiment if our intention is really to compare okay so we have these advantages of randomized experiment number one it allows the setup of um direct comparison between treatment of interest so direct comparison okay next minimize any bias so no minimize that anybody else why because we conduct randomization next is we minimize the error uh actually this is because of the experimental error or the the replicates now you you try to have more replicates or experimental error i mean experimental unit okay to minimize the error and then another is the is that the researchers are in control of the experiment again you should have a good background parameter control other factors that can influence the outcome okay okay so randomized experiment must be performed performed using appropriate principles and design and techniques of experimentation so we need to have this principles we need to know these principles okay so that we are confident with our experiment okay and we need also to know those design structures so that we are confident with our experiment to come up with a valid conclusion so what are those principles number one replication number two randomization number three local control so we start with this replication so it's needed for the validity of the experiment okay is the repetition of the application of treatments on a number of experimental units okay so repetition of the application of a certain level of treatment okay again um recall again that we try to apply a certain condition or a certain treatment level to several experimental units okay and then we expect that these experimental units must result to almost the same result uh measurements or values okay so that this actually ensures us that we are getting the correct result or results okay that's why making this experiment valid okay now functions of replications to provide an estimate of the experimental error so are you already familiar with this experimental error and replication also helps the increase of precision of the estimates and then increase the scope of the experiment so if you consider many experimental units and this experimental units have these inherent characteristics okay or conditions and then if we use them so we are also trying to you know widen our scope because right okay so that's the advantage now of replication another another principle that we have to learn is randomization is aside from providing the estimate okay of this experimental error you also need to have this estimate to be valid so if we have a valid experiment we also we also have to have a valid estimates so this can be done through randomization so it's a process that ensures treatments will be i will have an equal chance of being assigned to an experimental unit okay so we randomly randomize and then local control so is any technique used to minimize experimental errors although you want to measure experimental error but you want to minimize that okay we want to minimize the experimental error so that we can focus only on the differences or the discrepancies among the treatment levels or among groups okay makes the design efficient by making the test more sensitive and powerful powerful in a sense that we're able to compare or to detect any differences if that differences really exist okay so common techniques of local control number one use of loc most appropriate statistical design so you have to know those design structures use of proper shape and size okay then use of concomitant variable so this will help you okay although this is not necessary in your analysis but this uh you can try not to recognize them and control them and then we have this grouping blocking or balancing so these are common ways or techniques know to have this local control or error control so what are this grouping blocking and balancing so grouping is placing the homogeneous groups a homogeneous parental experimental units into groups and comparing the treatments in each group yes you can imagine we try to have this experimental units we group these fermental units now of the same characteristics or of the same condition okay so that's grouping then blocking grouping experimental units into blocks such that it's the units within a block are relatively homogeneous so we try to make sure that in a certain block all the units or experimental units inside there are of the same condition or homogeneous and then we also have this balancing assignment of treatments to an experimental use to achieve balanced configuration so as much as possible to increase the power of the test we want our design or experiment to be balanced meaning if we have this treatment one applied to four experimental units then we also have this treatment to apply to this another four experimental use of the four four so that's balance no but sometimes in our experiment it's very difficult to to acquire or expensive know some level expensive so we cannot afford to have this uh same number of experimental units experimental use for that expensive level okay but it's still now you can still analyze guide and balance now to summarize what are the principles that we need to know or we need to in the performance of experimental design or experiment we have what's that um randomization replication and local control okay so next is we have the steps in conducting an experiment so i'm not sure if you're already familiar since if you're doing experiment familiar nakaya if you're starting so at least this is your guide so state the problem the study formulate the hypothesis define the data that you want to collect or this define the responses that you want you know and then devise the experimental design and then conduct experiments so again before the conduct of the experiment you really have to consult your station oh what could be the appropriate design if i want if my objective is this if my research hypothesis is like this okay so before the conduct so hidden the other way around not from the not my data okay so next okay again experimental design concerns with planning experiments in order to obtain maximum amount of information from available source resources okay so we need to have a good land know of our experiments to maximize know your response and that involves the assignment of treatments to the experimental units so i think you already know that we have to randomize no we have this randomization of treatments now so just to give you a picture of this experimental design so take note that this experimental design has two structures first is the design structure the second is the treatment structure okay so yo so design structure is actually junior manga crd rcbd lsd so so that involves randomization and grouping while this treatment structure involves the factors so the number of factors or the combination of factors that you want to compare okay so a treatment structure consists of the set of treatments treatment combinations or populations that the researcher has selected to study and or compare so these are possible treatment structures as you can see in the right hand side one-way treatment structure factorial factorial arrangement with one or more controls nested treatment structure or fractional factorial fractional factorial so we will just focus on this one-way treatment structure but this is very i know uh important okay and then if you already understand this one way to construct you can proceed you can continue learning this other treatment structure the factorial or the fractional factorial or the nested okay now another component again aside from this treatment structure is the design structure so the design structure refers to the grouping of the experimental units in homogeneous group or blocks so this is where you okay as you can see this is where you can see this structures like crd rcbd and lsd there are also other structures like split plots triplet and strip light you take note that this split that can be combined with crd or cbd or strip plot with rcbd but we will just focus on this first three because these are the basic design structures which can also be applied later when you have this more complicated this unstructured exp split plot like those okay so try this okay so we have the same example the potato example i gave you wait okay this one okay so you are asked here what is the treatment structure okay since we are dealing with one factor with which has these three levels then we have one way as our treatment structure okay now what is the design structure although you have not learned yet this design structure crd but this is actually uh the randomization of different levels not to the different experimental units so as you can see randomized non-treatment applications don't sell experimental units that's why the design structure here is the crd okay but we will learn more on crd later okay and then another example is this again they have potato example perro hindi lang sol for application concern we also have this other factor the sprinkler irrigation use so which can have uh this center pivot wheel line and solid set so we have these two factors and then we combine these two factors so that is actually factorial now so the treatment structure is factorial but then the design structure is still crt okay since this different combinations of different factors are randomly assigned to the different plots okay so maybe we'll have can we have a break or yeah i think we will have a five-minute break at this point and marty yeah so uh okay we will have a break so in terms of for today so if uh [Music] uh discussion imam so we'll give mam lara a mini break uh so before uh while we are on a breaks my my idi discuss what i see sir jeff okay so okay we'll talk about our collaboration project with the at the nato chemistry society so this actually focuses on uh the creation of uh stem teaching modules particularly in chemistry biology and physics so um marty can you see my screen yes so and then what you do is um you scroll down and when you scroll down go to the bal um click on the button so this is build a lab this is a collaboration project between the athena native chemistry society and filipino science hub so what is baal sobal is a year-long flagship program of the ateneochemistry society so historically paul what they did was um talawa at least high schools within in manila so they provide resources for um for them to be able to perform on laboratory experiments so we provide equipment they provide uh reagents for for chemistry experiments so don't always but then since we are so since 2020 we we got surprised by the pandemics for their program so specifically what they did was um instead of delivering equipment not great boosted and grade nine chemistry modules and so what they did was they approached us uh to partner up with us because since in addition to chemistry mode our teaching modules we have also been coming up with a biology and chem biology and physics teaching module so example and we combined all of the contents on the bal web page which again you can find on our official website so go to triple w fills ihub.com so it has all the information about the ball project this is actually like a very good project that ateneo is i think chemistry society has been uh generous enough to put up for this year to our chemistry module so what you just have to do is you have to click it and then if you go to that page it's build a lab chemistry modules uh tab and it is models that we have um uploaded so the first of which is the worksheets chapter it covers a number of different i mean i will click on it so you can get an idea of what what what each content um has so this is the specific worksheet grade nine worksheet that uh at the neochemistry society had actually composed so this is for grade nine specifically and then if you look at the topic so it's covering several chemistry topics for high schools electronic structure of matter chemistry chemical bonding organic chemistry mole concept special discussions um chemical formula nomenclature dimensional analysis and what's more important about this is that neter and pusilon worksheets so i don't practice problems for students so and these are things that you that you can actually access through the val webpage okay i'm going back to that page because uh chemistry uh uh modules um for for the pal program is if you click on these links it will direct you to our google drive um page where you can directly download everything as pdf so your documentation downloading teaching modules for chemistry so for example for ideal gases um what we provide is uh the set of core standards course objectives and at the same time you know like some of the most essential learning competencies that should be taught you know um for these lessons with the guidelines of them and then we begin with a review or a reconnect of some of the lessons related to to each topic we provide um sample problem solvings um um for for for for the concept so it matters on discussion um um problem solving as well and then uh at the end of all these laggy putty on and then in the same package there are also um some answers there is an answer key that is provided so again lecture part or teaching part um like i said we have that for chemistry we also have that for biology so focus [Music] and these are all um teaching modules and and problem sets that we created and curated specific to high school um chemistry so both junior and uh senior high topic so ibrahimos if there are things that are relevant to you in your classes so these are complete um teaching modules and then last people we also have our physics teaching modules so so far we have about eight teaching modules on specific physics topics so again same content it will have uh the the course at the teaching guide it will have the discussion part so you can use it you can get some visual aids from these packages and then um lastly again we have the problem sets and then uh what's one very important thing that we would also like to uh to mention is the fact that um very recently some of these modules um we uh came up with audio visual versions of these so ponta po is a youtube there are there are some discussions somewhere upon video discussions um guiding you through each module again this is our bal um project in collaboration with the ateneo chemistry society and it's uh so our webpage is already live um you know for for more than a week now okay and if you have questions about the the ball uh modules are reach out to us through uh the comment on our websites and um and also reach out to us through our facebook page thank you for uh that uh discussion so uh welcome it'll be welcome edition partnership nang phil's other partnerships to launch uh with many schools and many organizations so watch out for our partnerships and launching of other new programs so mam lara are you ready to proceed yes paul i'm ready i was sharing about my slides my next set of wait it's here okay so can you see the slides now yes ma'am can you see me yes can you hear me yes good i'm good so again so we'll just continue on with this discussion on experimental design so we so far we learned about those um terms that's very important terms now so for us to at least uh uh young confidence nathan's uh experiment in doing the experiment okay is matas okay now the design structures that i uh i was talking about okay uh that will be discussed this uh for this webinar are only this uh crd rcbd and lsd so we'll start with the crd so what is crd okay so if you can recall i mentioned kona you see arduino so this is just a completely randomized design so this is the most basic experimental design okay all experimental units are considered the same and no grouping among them exists so there's no grouping okay and then the treatments are allocated randomly to the whole set of experimental units so that's the eu's and each experimental unit has the same chance of receiving any of the treatment so as you can imagine so we have this for example different uh plants okay and then we have this um treatment levels blue orange and green so we apply the blue okay treatment okay or the blue fertilizer for example so to this plant so la random lang so we just randomly assign so and then also for the orange okay fertilizer okay we can randomly apply this to the same experimental area so we have this plant here receiving orange treatment and another plant receiving the orange treatment and also for the green okay so this green uh plants actually receiving the green fertilizer okay so just a picture of this crd now take note that okay they are used when experimental units are homogenous okay meaning the same the same variety the same age okay so depending on study search or experiment okay and there is an effective local control okay so how to have this effective local control okay so how to minimize experimental error so that we have this effective local control so usually this crd is usually done inside a laboratory or greenhouse okay union effective local control inside the laboratories because we um there is the same temperature or same set up like that and then we have this requirement so take note of this requirements t treatment levels to compare so take note we have three or more levels that you want to compare and then you have this n experimental units must be homogeneous again homogeneous and then n must be greater than equal to the two two times the number of treatment levels so if you have three treatment levels so two times three you have how many at least six okay six um experimental units okay and then take note also we have this ri where replicates are assigned to each level of d so for each level there should be a value of r so for level one you have r1 for level two you have r2 and so on so now this replicates okay can be can have okay can have the same number okay we have r1 equal to r2 or equal to rt okay so we can now have a balanced design now for an unbalance so you cannot sometimes afford to come up to have more replicates for a certain level so we also have this unbalanced design or experiment now since we are dealing with randomized experiment so we need to have this randomization so how do we uh generate or how do we conduct randomization so first okay we have the steps assign the first r1 to t1 next r2 to t2 and the last r3 to t3 okay now second obtain a sequence of random numbers so you can use clear calculator or excel okay or you don't need a software for this no wedding third step is to rank the random numbers okay so as you can see here in the illustration having three levels t1 223 and take note that p1 okay is assigned to how many replicates two so we have t1 t1 here and then t2 is assigned to three replicates so we have three right that's why we have t2 t2 t2 and then for t3 we have four replicates so we have 43s okay now we generate random number okay so we have this so it can be any numbers no from zero to one and then this uh next is we get the rank okay the rank okay so the lowest generated random number okay this one here has round one and then the second to the last lowest which is point two one eight has rank two okay so these are the ranks now and then we can have the layout okay so as you can see t3 is assigned to this one here and then t1 can you follow the for this two we have t1 and then for this three we have t3 again then four we have t2 all right so this is now your layout listed and now next yes we expect to have this data presentation so since we are dealing with three or two replicas under treatment one then we have two observations okay because we have two replicates so from each replicate there is one observation okay so for treatment and then we expect to we use three replicas then we expect to have three observations so this is the data presentation then we also have this replicates the number of replicates and then the total and then the mean so basic data presentation and then take note also that once we have the responses we can model the responses so we have this means model and we also have this fixed effects model so for the means model we have this y i j equal to mu i plus epsilon i j so um this y i j is your observed value response and then the mu i is the mean of the i treatment level okay the mean for that group for the treatment level and then the epsilon ij is the random error okay so we can also okay have this effects model okay for the responses so instead of having just mu i plus mu i plus epsilon we have this mu plus tau i so this mu plus tau is just equal to miui so meron so upper uh right meron tayo wayno to come up with this um okay to relate this model to the previous model okay so in the previous model meme ui so this mu is just your mu plus tau i and then tau i now is your mu i minus mu this is used okay this model is used when you want to estimate for the treatment effect so the taoise that's it and then the mu i j again your response or the observed value and the mu is the overall mean the tau i is the effect of the ice treatment level and then the epsilon is the random error so that's it okay this is very these models are very important in stating for the mean and later on the conclusion now okay so we will now have this example under crd okay so an experiment was conducted to determine the effect of levels so we have 30 60 and 90. so this is very common example actually of nitrogen on the yield of variety of corn so 12 homogeneous plots so imagine the plots okay so 5 by were where prepared and each level of nitrogen was randomly assigned to four plots so merengue three levels then you randomly assign these levels or applications to the different plots okay now okay recall treatment we have treatment is the amount of nitrogen that's 30 60 and 90 so that's the level in experimental units plus in applying nitrogen and then the response variable the yield of coin okay so this is now your data so it's a balance now we have this question at alpha 0.05 are there significant differences in the mean yields we're talking about comparing the means mean yield among the different nitrogen levels considered okay so how do we answer that okay so we have to recall the steps no in order to answer that question okay you have to take note state the h o the h a decide for your alpha where in here we already stated that alpha is 0.05 so we set alpha equal to 0.05 select the test statistic so you have to know what is the upper test procedure so that you will be able to know what is the appropriate statistic and how do we compute them okay and then collect the data okay and make a decision then write the conclusion okay so start with the hypothesis okay as i've said meron fix effects random effects so this will be reflected though and random effects will be reflected in the hypothesis testing but you can actually always use this means model now we just use the mu so for the null hypothesis using the muse we have mu 1 equal to mu 2 equal to mu t so meaning so in words you say all the treatment level means are equal so the same langu means now from different groups or from different levels versus the alternative that at least two treatment level means are different so or at least one is different from the rest so this is somehow a friendly way in stating the null and the alternative hypothesis but if you want to know what could be the hype the the null or the alternative hypothesis when we have the fixed effects model then we can have this tau one equal equal to zero and that would mean that there is no effect at all of all the treatment levels so actually they just have the same meaning you means model fixed effects model and then what did they use a random effects model wherein you use this notation sigma squared tau equal to zero wherein there is no very short but no variation so there is no variation among treatment levels yeah so i am just showing this to you just to give you an idea and uploading um you state the h on the h a the other way around you aside from the means you can use the fix or the random effects okay so go back to the example we state now okay using the mu 1 mu 2 mu3 okay we have your h o here or we say the mean yield of corn among the different nitrogen levels are the same versus the h a at least mu1 mu i not equal to mu j where i and j okay i or j equal to one two three okay and then at least one of the nitrogen levels has a different mean so you know ebx mu i not equal to mu j at least one nitrogen level has a different mean okay so the test procedure i i don't know if okay familiar numbers no okay so one procedure is we can use the f test using one-way anova bucket because say if we satisfy the assumptions we can use this test if we do not satisfy the assumptions we cannot use the f-test so we have to go we have maybe to transform the data or use non-parametric method okay so so not all the time that we use the f but this time we will just focus on the f and then also given that the assumptions are satisfied okay the test statistic since we are using the f so we have this formula this is the ratio of mstr so mstr is the mean square of the treatment over the mean square error okay so how do we get this value so we need to construct an anova table okay then decision rule we reject h0 if the f computed is greater than the f tabulated so always greater kappa that we reject okay and then okay this now the the right hand side is your okay the f tabulated okay if you do the manual okay way of the analysis um we we can have this um f tabulated okay excel or we can use another way of deciding whether to reject the h o is whether or not to reject the h o is to use the p value so hindi f value so reject h0 if the p value is less than or equal to alpha wherein we set alpha equal to 0.05 in our example here otherwise fail otherwise fail to reject the h0 so you know ft are here okay so again as what i've said this anova is very helpful in computing for the f which is the ratio of the mstr over mse so what is really anova so anova is a way of partitioning the total variation so the first partition is the variation due to treatment which is actually computed using the trss which is anova table and then the other partition is the error sum of squares okay are actually computed because we want to have this variation due to randomization so reflexa as you can see in the anova table the ess okay now we start from getting this sum of squares and then we divide it to the degrees of freedom to get the mean so from the total we get now the mean and we get the ratio of this and we get the f and that's it you now have your x value okay but again take note that we cannot use the f okay test if the assumptions are not satisfied so what are those assumptions number one homogeneity variances we need that number two normality number three independence of experimental errors so this time i will illustrate to you on how to check on the and these assumptions okay i'll start with this um visuality of variances so we have homogenized variances so the variances of the experimental errors of the groups must be the same so you expect here for example you have how many applications of is that sulfur or nitrogen so we have three so you have three groups okay for group one for level one or the nitrogen level 30 and then the next group okay group two and then the next group group three so you expect that we have this different variances from each group and that should be uh this biases should be the same okay so we have to test that so that's rho so sigma squared 1 so this is the variance under group one okay equal to sigma square two this is the variance under group two okay and so on okay so this is a very general hypothesis but in our study we only have until sigma square three sub three so in short we have this error variances which are equal for all the treatments versus the h a at least one so it is from the rest so at least one pair of error variants are not equal okay then we have this test procedure levine's test or bartlett's test these are common tests okay you can use r or you can use um sass or you can use data not to come up with these values and the decision rule okay we may use the p values if we're using a software say p value less than alpha otherwise uh reject h oh if p is less than alpha otherwise failed reject h okay so this is an example on how we uh check for this assumption um we have the star software i'm not sure if you're familiar with star but this is a free software wherein you can download and then install it for free okay you may go to erie website okay and then okay later i will illustrate to you how to do this okay um we have this p value this is very important so this will help us decide whether to reject or fail to reject now if this is greater than what is our set value 0.05 so if this is greater than 0.05 then we what okay fail to reject h0 and it's a good news to us because meaning the assumption is satisfied okay so if it's great if the p value okay is greater than alpha we fail to reject the h o okay and that we can say that the homogeneity of error variances assumption is satisfied and that's good okay another assumption that we have to deal is this normality of errors okay now we have this um experimental errors the different groups must follow a normal distribution okay some popular tests we have the sky square called mcgruff smirnoff and then we'll shop at all okay but uh young common detail is shopping milk or wheel okay so ho here stated as the errors follow a normal distribution each a the errors do not follow normal distribution okay then we can use the software the star software to check on this okay uh the test using sharper wheel and then decision rule reject h0 if a p-value is less than alpha otherwise fail to reject h okay an example so since we have different groups so we have to check a normality for each group okay so we have p-value okay for group one we have point ninety-two that's greater than way higher than point zero five also point ninety-nine way higher than point nine eighty-nine so good news for the hill we have satisfied this assumption of normality so at least we can proceed okay after the homogeneity variances we have this normality of errors what's next the independence of errors so independence of errors are actually okay must be part of the design in terms of randomly assigning the treatment so if you if you're confident that you have randomly assigned the treatments to the experimental units then you have satisfied this assumption of independence of errors okay and then okay we go back again to the example okay so we have this mu our h o and h a so i've discussed this already the test procedure the f and then the statistics we have this ratio and then the decision rule you can use the f and the p value okay um if for example you don't have the software or you are you are not uh if you want to at least try no doing the manual computation so you have this formula here so you are provided with this formula okay and then for sure this formula will have will give the same result with that of the software okay so okay just to show you know my correction factor my total ss my treatment ss and the error ss okay okay now this is our anova table so we have this ss take note we computed for 9.4 2.66 okay and then 6.74 so these values are found in the next slide to 9.4 yes this one this is the total s the tss the ess and the trss so this one divided by the degrees of freedom then we have this amine square and then we get the ratio and then we get the f and then we compare so how do we compare so we compare it to the tabulated so i have i i'm showing here the formula or how or the function and how to come up with this 4.26 so you can use excel f table you have this formula i'm showing and then just press enter so input this okay and then enter and then you will have this 4.26 and then you can now compare so 1.78 is less than 4.26 so we that means that we fail to reject because we only reject if it's greater than and sadly in this case we have less than so we fail to reject and therefore at alpha point zero five the mean yield of corn among the different nitrogen levels are the same okay so paraslam don't know result or you means no different levels okay so if you are a researcher or if you are uh trying to compare this and if you see na they are just the same then you just go for the less or the least expensive level okay another okay this is an output coming from this r okay our software uh we have this p value point two to say so we can use the p value to compare i mean to decide whether to reject so point 22 is we compare this to point zero five okay that's greater than point zero five so we fail again so consistent okay consistently result not enough regen uh using the f and using the p okay so the same down long new means okay so how do we okay i'm introducing now the r software because it's free and then you can just download this so you can go to this uh to this link onto this website okay and then download this two so this star will not work without this r package 1.2 so you still have to download our okay but it's fine because after this it's easy for you to do the analysis now just to show you and how to use star okay using uh to come up with the layout of this randomization so they have this uh feature here we have design and then completely randomized design and then you just set your values here so you have four replicates and then the number of field rows so i want the layout to have three rows and then just one for the for one trial and then take note also of these levels you have three levels so you have three here then okay then you have this result so easy you don't need a calculator or an exam no you just have to use this star and then come up with this layout and then you have the idea on how to randomize or how to randomly assign your treatment levels to the experimental units now another is when we do the crd okay analysis here so before the analysis of course you have to input your data now so you start from project the new project and then you try to write the name the project name and then okay and then from that automatic there is this folder here crd with the name of your project and then there is this folder for data and another folder for the output for the data you just right click and then you can see the import data and then you just choose data negotia analyze and then open then you have this so easy so you can see the data if you click the data folder here okay next is okay how to do the analysis okay so once you have inputted the data okay so you go to analyze then analysis of variance given that assumptions are satisfied here okay so there is this feature also in checking for the assumptions you can go to this descriptive stat here and then have this um normality test okay so this one i'm showing to you is the yeah given that assumptions are satisfied so you have this analysis variance in the crd and then you just click this tab tab here and then the tab will show this uh you have to um move this uh variable name so yield okay that's your response and then the treatments that's actually your trt based on the data that i have and click ok and then automatic you have this anova table so easy okay but take note again before doing this you have to check the assumptions okay now think about this knowing that at least one of the treatment levels has a different mean how do we know which treatment uh mean differed significantly from the other treatment meat so meanings how how do we know that which which level or which group is different from the rest okay so anonymous so the answer is by using pairwise mean comparison or group comparison so for the pairwise we expect that we have this for example three levels compare a with b a with c b with c and so on and then for the group comparison you can compare group a with group b and c together or group b with group a and c together so that's under group comparison okay we will not be um discussing this because somehow there are there are lots of procedures under this okay okay but at least you have the idea that you can proceed no to pairwise comparison in case the you have a significant result now another thing that you have to think is in reality it is very difficult to find experimental units that are homogeneous i hope you agree with me right especially if you're in the field no and then experimental units are markedly heterogeneous with respect to some criteria of classification so we call them um nuisance factors or criterion so example plot may differ in terms of fertility although your intention is to compare just the the type of irrigation okay and then you apply this types of irrigation to to the plots okay but then this plot has this inherent condition or this factor okay that they have this um not the the fertility of this plants are not uniform okay so you cannot control that okay another example is trees may differ in age for example you want to compare the different levels of fertilizers applied to trees but then the responses or say the harvest or the yield of these trees okay may differ or may be affected with the age of the tree right so that's another nuisance factor that age another example soil may differ in terms of ph or acidity okay so when you want to compare the soil maybe in terms of level of the fertilizer okay so i can have this another nuisance factor the ph that can affect yes can i just interrupt i think the the slides are uh i think stuck because we are only seeing this uh pairwise mean comparison and maybe i have to refresh uh can you you could stop sharing for a while okay okay it's okay okay so i'll go back to the slides where is okay i will try again oh it's that sweet okay so you're you're until here yes yes and then if i move this next slide yeah okay yeah i hope you you understand my point now that there are other factors that that's already there in the experiment or in the research that maybe you have to consider those and then you have to control it no okay so yo so these are the examples so do i have to go back to these examples oh it's okay uh i think factors because another last webinar okay another thing that you have to think okay this one yeah you sense factor it's a factor that probably influences the response variable but it's not of interest so how do you consider that okay so we now have this um blocking okay and also have this um another statistical technique that we have to use okay so but we will only okay we will just discuss on this rcbd which is actually under blocking and then that's a square design another uh procedure under blocking okay and then this incomplete block design this is another design but we will not discuss this since but at least the basic designs or the structures the sense structures are okay we'll be discussed later so under blocking now for the use of another statistical technique we have this anakova or analysis of covariance so what is this oh this is just the a procedure with the combination of regression analysis and anova okay so that's very interesting actually but we will just focus again with the rcbd and the latin square so now i'll start with the rcbd okay so rcpd are called randomized complete block design okay this assumes a single gradient running across the experimental units so it is factoring consider nothing that single gradient or we call it block okay so the blocks or the blocking factor blocks are of equal size and each contains all the three the t treatments okay so imagine that you have one block or a block having all the treatment levels there so my presence from the heart no treatment levels that's why it's called complete block okay so complete block design and then the experimental units within block are homogenous you expect that in one group or in one block the conditions for example the variety or the type of soil okay under homogeneous condition experimental units and then the variability among blocks is taken out of the experimental error so because our goal is to minimize experimental error okay we need to consider a blocking factor so that okay will be okay incorporated in that blocking factor in a source of variation okay but the goal is to really minimize the experimental error by having this blocking factor okay just to illustrate so this is a picture of a single gradient example a fertility gradient so as you can see uh since you go to the first row uh les fertilla compared the last row okay can you follow okay so for young level now fertility across rose okay since if you have the idea so that's the situation of the experiment so you have con you have to consider blocks here wherein you have block one having the same level of fertility or the for example black one having uh less fertile no part and then the block three with the most the most fertile part in the same block okay and take note that block one has this three treatments okay present and block two the same the three treatments are all present and then the block three also has this three treatments present okay so now gradient is unidirectional okay this one one direction long orient block such that they are perpendicular to the direction of the gradient so if this is the direction of the gradient so you're going to unblock small and then use long and narrow blocks so this is just a suggestion okay to accommodate all the treatments so we have long and narrow blocks okay oops so we have we now have to consider uh to consider this now considerations in blocking first select source of variability as basis for blocking factor so you really have the idea i know you possible factor a nuisance factor no okay in the experiment so that's your blocking for your boy then select the block shape and orientation that's why now we have perpendicular in this shape long and a row and then we have to consider group uh grouping of these experimental units into blocks such that the variability within each block is minimized okay you discrepancies within the block is minimized so hello okay while the variability among blocks okay are maximized so teletunnel dd first in black one for example in soil type soil type clay that you definitely saloon and so on okay so we have this application so usually human applications for is uh on this agricultural experiments because basically you experimental design actually was developed now with this uh from this agricultural experiments okay we can also apply this um rcbd sorry uh to biological or medical experiments so think of something uh experiments on biological medical for example you want to compare the uh say different dosages of a drug okay and then a new possible block could be in the experiment that can also affect nonsense effectivity now treatment it could be the age right so your age can be your blocking factor the market research or in business for example you want to compare the different brands you want to compare the different brands okay and then so you price as you go your space variable and then young blocks could be the different department stores or different malls okay and then engineering industrial experiments uh maybe you want to compare the different um temperature okay different levels of temperature and then your block would be the different brands of machine okay so that's it so those are some applications of using our this rcbd so randomization and layout so we have this illustration so electric led treatments t1 t2 t34 uh be replicated four times respectively thus the total number of experimental units needed is 16. so if you can imagine we have uh three uh four treatments and then we have four replicates okay so these four replicas actually on four blocks so one block one replicate having these four treatments another block another four treatments another block another four treatments another block another portion all in all we have 16 so that's four times four now we have to group these experimental units into four blocks okay so you know do the grouping okay okay and then divide each block into t uh experimental units okay so one block four experimental units since we have four treatments so for each block randomly wait we cannot read we have to move randomly allocate the treatments into the experimental units at random okay so take note that uh hindi uh directna in applying the treatments you really have to randomly assign these treatments to that block or to the second block okay so now we have this data presentation so if you have this four replicates so we have one two three four or four blocks and then four treatments one two three four so we have this so 16 observations and then we get the total or and the mean and that's all okay so so that we'll be able to understand this distance structure so we have this example an experiment was conducted to determine the effectiveness of four weight reducing programs so there are four programs each program was randomly assigned to any overweight individual of each age bracket and the weight loss in kilograms after 30 days was then recorded okay so just accounting recall uh treatment here is the weight reducing program and then the treatment levels independent it could be uh level one two three four or abcd and then experimental units we have 16 okay individuals then response variable is your is the weight loss okay so it's blocking needed yes okay why because age can have an effect on the weight loss right okay so you have blocking factor four age brackets okay now determine whether at least one weight reducing program resulted to a different weight loss mean so also determine if the blocking is effective so you have to test two for the treatment that for the program and for the um age bracket or age group okay conduct at uh conduct experiment at five percent level of spread of significance so now to illustrate okay we go back to this randomization panorama randomly um assigned no non-treatment so these four programs so okay this age brackets are already set so fixed okay blocks are fixed so we have 15 19 20 24 25 to 29 30 to 34 years old and then for this first block we assign uh we have this okay anupan except munateng numbers one two three four five so one on your arrangement ascending order and then the next step is to generate random numbers so if we generate these random numbers okay take notes for block one what is the lowest value here so it's point two zero nine so that has rank one or rank one and then the next is point three six eight that has rank two and so on so gonna then you have to assign rocks again for the next block okay having rank uh having this smallest or generated random number rank one rank one and then the rest okay so i hope you understand you can follow this one and then next is we now assign the treatment so take note that the rank one okay okay of this experimental unit will receive okay this treatment a then two b and then three c and then four d then the same process in the next block you can have this one as treatment a two as treatment b then three as treatment c then four statement d and so on so i hope you follow okay now okay so given that we have randomly assigned those treatments or programs not to those individuals overweight individuals and then okay so you you apply this program and then you wait for uh 30 days okay then after 30 days you measure your weight yeah so your weight loss okay as a diet program and weight program weight reducing program so you abcd and sodium specific okay we have a intermittent fasting bee vegan diet vegan diet low carb diet ultra low fat diet for the day and then we have this result and then how do we analyze now okay so take note that this um out responses here or the weight loss actually can be modeled okay using this okay so recall before in crd we don't have this term the raw this one the third term we don't have this we only have mu plus tau i plus epsilon lj now for the rcbd we have this additional term okay if you're just interested not to model the responses you have this now y i j is the weight loss or the observed value the mu is the overall mean that that was the effect of the i treatment and then the raw j is the effect of the block and then the epsilon ij's experimental error okay so again we have to check first the assumptions before we proceed now with the test okay so assumptions on this last time we have discussed this um homogeneity of variances the normality of errors and then this independence no another or additional assumption that we need to consider is we need to actually check is the additivity of these treatment effects okay okay so this is very important okay you cannot uh proceed with the crd or the analysis for the crd if this um assumption is not satisfying okay so what is additivity of effects so treatments and environmental effects so you environmental effects are you new missus factor or intuition blocking factor so the treatment of the blocks okay do not interact so that pathway interaction meaning the effects of treatments across blocks okay effects of treatment so treatment one sublux one treatment one's a block two treatment once a block three and so on are uniform and the effects of blocks okay across treatments are also uniform so in short well an interaction specific combination of treatment and blood now that has an effect result okay that's uniform and then we can test this using the two kiss test okay so two keys test for non additivity okay take note we also have this ho the treatment and the block effects are additive or do not interact and then we have this h8 the treatment and the block effects are non-active additive then test procedure took his test for non-additivity and then decision rule okay we reject h0 if the p value is less than alpha otherwise failed reject rejection so we use the alpha here but um okay if we are using the software okay so just to show you the result okay so using two keys we have this p value point eight so this very high so meaning to say we have satisfied assumption of additivity and then you if we want to picture uh to somehow at least have an idea of the picture of this non-additivity you have this graph here roughly parallel okay so this will give you a hint indica test this will give you a hint that we have this uh additivity assumption satisfied since parallel same flow or throughout no all across all throughout the treatment levels okay per block same length behavior okay so that's one way then another is to test again for this the formal test is to use the two keys okay homogeneity of error variances we're already familiar with this so the same h o k sigma one square equal to sigma two squared and so on and then the error variances are equal so that's your hoha at least one pair of error variances are not equal then the test procedure are the same levelings are bartlett's and then the decision rule will reject h if the p value is less than alpha otherwise fail to reject h okay so this is an example okay we have this levine's uh result test result the p value that's greater than alpha for the treatment and even for the block 0.88 it's still greater than alpha so therefore we have satisfied the assumption of what the homogeneity of error variances okay the next example the next assumption that we have to consider again is the normality okay so you have this ho follows a normal distribution h a follow does not follow a normal distribution and then we'll shape it or chop it away we say about software then we use the p value so assumptions are per group okay okay so we see that the p values you can see here point eight point eighty five point five one point this is all of this are greater than alpha which is point zero five therefore we have satisfied assumption of normality now the independence of error just the last assumption uh satisfied okay again using randomization okay so we have the summary here so for that satisfy so we can proceed with the text okay so the test now is this f okay so take note now aside from the muse that we usually use okay we can have this fixed effects now h0 for fixed effects and then h o for the random effects so then tau i and then sigma squaring now your notation is nothing and then the f again go back to f we have this ratio and then decide using f or using the p value okay take note also here okay here we are dealing with uh with the comparison among the treatments right in terms of means and then here we are dealing with the comparison among blocks so we compare blocks so we have to test h0 blocking strategy is not effective each a blocking strategy is effective okay so we have this so but not effective no differences in the means among the blocks okay so to summarize okay we have this a hypothesis for the fixed effects and then random effects okay and then we have this test statistic the numerator under the treatment is the mstr now under the blocks we have msr is actually replicate or block okay and then the decision rule you can use the f or the p value so i hope you can follow so pattern with let's take the ho the h8 and then we should know the test and then we check the assumption and proceed okay then decide and make a conclusion okay we go back to this example okay so we have this i am showing you the the manual computation okay in case long you have time to to check or to verify now same like your results software and then you saw manual computation okay so we have this okay anova table here so take note that there is an added variation this is the block as the source of variation before for crd we only have treatment error and total now we have this additional variation okay so we have this row okay and then using the i'm using now they are okay our software to come up with this um uh result okay as you can see significant highly significant young treatment but for the block which is the age it's not significant okay now okay so this is just uh a quick review so we have to state our h oh here no differences among the means each is there at least one that program has a different mean and then test statistic we use the f okay assuming satisfying assumptions and then decision rule here we have this and then recall that our p value using the software we have point zero zero four this one here so we can now compare this to the alpha point zero five then since it's less than alpha we can now reject that h0 then the conclusion we have sufficient evidence to say that at least one diet program has a different uh weight loss mean or the diet program is effective okay effective okay now since okay before uh in the previous discussion about subpac uh if in case significant tune treatments or young testing treatments if okay there is this significant uh test uh or we found out that there is this uh difference in the means okay we cannot proceed to pairwise comparison so for the pairwise comparison using two hsd this is automatic when use the um r commander so our software okay so you can uh use this um and then you can check the p value so the p value here for the b and a so pair b and a has this point zero zero two meaning to say this b and a d first in terms of their means right okay so this diet program a and diet program b differs and then another pair that differs we have the c and b okay so this b again and the c and then another is the b and the d so therefore in short we have this diet program b that differs from the rest i hope you agree with me uh we can go back to the raw data i'm sorry we have if we have if we go back at least we can relate now back at the new compared to the rest of the diet programs okay so at least now we understand we can relate the result to that to the data that we have or we can even get the mean the descriptive statistics where are we now okay so we're done with this okay we have tested for the treatment means and then we have tested for the despair wise comparison and then now for the blocks okay now the ho the locking strategy is not effective so this is denoted by raw subject equal to zero versus the h a r sub j not equal to zero or the blocking strategy is effective okay the test statistic again we use this ratio where in the numerator is now for the block so msr the decision rule as usual we reject h0 if the p the f is greater than the f tab or the p value is less than alpha now we go back to the anova result oh see that three two 1.1322 okay this one here i know various other point one three two two is actually now compared to alpha point zero five and then okay that's greater than point zero five that value point one so we cannot reject the h oh okay and then our conclusion now okay the blocking strategy is not detected okay or there's no sufficient evidence to say that at least one age group has a different mean okay so the blocks are just i know of equal means okay i'll just show you i'm just showing you this uh code no if you are using the r okay software so this is very helpful especially in getting the result for this two keys now for the id for checking for the additivity assumption okay now we can now compare the efficiency between this rcbd and this crd okay so to measure the efficiency of rcbd with that of the crd we can use this formula okay and if the result using this formula is greater than 100 then this implies that rcbd is better than the crd okay okay and then if it's less than 100 percent this implies that crd is better than the rcbd with the same number of replications okay so let's have an example under this um recall the formula or the anova result that we have in the example um in the formula here we can plug in the values okay so this 0.7456 is just your mean square for the block okay so that's point seven four five six and then another value point three zero six seven is the mean square for the error or the mse so we plug in the values and then we get 128.62 percent so lamp 100 therefore there is uh this gain in efficiency in using the rcbd over the crd with the same number of replications so rcpd is 28.62 more efficient over the crd with the same number of replications okay now to be equal in terms of efficiency okay we need to what multiply this value 1.2862 times the number of uh what's this replicates the blocks here the four okay so we have around six okay replicates crd okay we need to have six replications okay to have the same efficiency with that of this rcbd okay so what if experimental units exhibit heterogeneity into directions of classification so you have to think uh in the rcbd we only consider one blocking factor how about now if we consider two blocking factors okay this is the question now the answer is use two-dimensional blocking factor or we say we call it latin square design okay so do we need a break now if i'm not mistaken it will be young design of experiments between crb bc cr completely randomized crbd or rcp rcbd that was after that you will have to test for your assumptions before you proceed testing your statistics you you should ask you should test your assumptions first so maraming after that you uh you test for your blog statistics [Music] sir jeff to have some announcements sir i would just like to point out no in the interest of typo [Music] it's to introduce you to all the statistical tools that you can use for designing your experiments so you just have to try out you just have to start with the design of the experiment and then figure out what would actually be the most suitable uh statistical tool that you can actually use in designing your experiment and something just like any mathematical tool um you level you you you confidence you possibly come with the motto it comes with practice and then all of a sudden expert now so you have to apply this whenever you are um working on a research project and um there's a need for you to um adapt uh statistics as a tool to design your experiments you take that chance and then again at the same time if you get lost because just try to to reach out to people who are experts in statistics because i bet you know i'm in your local universities um there should be statisticians out there who could help you out okay so um i would like to give you uh guys an update on the upcoming events uh that we have um for the months for the remainder of this month and also extending over the months of may and june okay so you guys can see my screen right all right so today paul we have miss lara um it's it's it's our training course and music statistics and next week so those are teachers and students are actually quite interested in the area of cancer research molecular biology and and the area of medicine we have um professor chad creighton from the baylor college of medicine so he's actually working for one of the leading research institutions in the us when it comes to to medical research so he'll talk about cancer genome analysis so ital actually this is more more biophysical in nature coding so if there are students and teachers out there who are interested in um [Music] how you can actually use like um um computer science you know specifically um applied due to cancer research this is actually one thing that you should watch out for and then um april 30 end of this month is the deadline for submission of entries um to filipino filipino science of research universities um lions then competition so this is a competition uh where teams of teachers and students are actually invited to submit a research proposal so essentially come up with a research proposal for a science investigatory project and what we will do is that we will screen all those entries and we will pick uh the top 12 finalists um from you know from entries all from all over the philippines and then what happens is that um the semi-finalists will present uh their ideas in front of an esteemed panel of judges um these are seasoned filipino scientists who we call our lions and then we will be picking four winners here so the top prior price so the winning idea will receive ten thousand pesos for and that you can the teachers and students can use it um as seed funding for for carrying out their research project and then we will also be picking three uh lions choice so merlin coming to brothers up each of which will receive each group will receive about 5 000 pesos in prices for and then another perk is the fact that all the finalists who will be chosen will be mentored by filipino scientists from different stem areas mentorship from a selected panel of mentors so we invite you all to participate in this again the deadline is april 30 2021 and if you have questions about the details uh we have all the information on our facebook page and also on our website so just look uh just look for um phil say hub research university there's a menu right there and you can actually find the mechanics and guidelines okay and then after that um i believe may 8th right marty and jp we don't have it here may 8 we do have a special topics webinar on um um computational chemistry so research topics that students and teachers can actually pursue at the conference of their home and so we mentioned computational um science or computational chemistry and and modeling and in that talk and so the request for io no computational chemistry um um topic and we're gonna be delivering it so this this particular webinar will be delivered by professor ricky nellis from the university of the philippines the le mans so is he's one of the leading uh experts in the country right now when it comes to computational chemistry and modeling um study so may 8 watch out for that and then the week after that may 15th will be the one of the mandatory courses for filipino science of research training program so this is our uh training course number six which focuses on how to write a scientific report so when you're working on a thesis or when you're working on on a research project in high school this is actually a lecture that will uh give you um an idea of the fundamental parts of a research report so whether my publisher says um um scientific journal or whether max is a bitcoin report as output um um high school teacher this is actually a training for you so we'll not only talk about the fundamental parts of the scientific report what i will also be doing is that i will try to [Music] essentially summarize my 15 years of experience of writing different forms of scientific reports you know like from from project reports i'm being submitted to government laboratories all the way up to a scientific journal and patent writing so that i will be sharing there will also be some some best practices and we'll try to cover some examples then um in this in this uh our training course and then on january on june 5 of 2021 we will have another web lecture in organic chemistry and um this time around professor dundee you of uh phil sci hubs a head of international programs and also professor of chemistry in louisiana university she'll introduce you to the world of alkenes and also in that particular lecture i will be introducing uh the high school teachers and students in even those and at the college level who are quite interested in the like in the topic of reaction mechanisms um organic reaction so we invite you all to you know um sign up for these events these are all for free um take note of of those dates and i think um on june 12 we will also have a green chemistry uh webinar and that is to be delivered by professor cat vasquez from i think the university of santo tomas so um events um um over the next couple of months so we invite you all to participate and sign up in advance um we're looking forward to having you in all these events all right so thank you huia jeffs exciting events after this and of course this event actually is uh well uh received by our audience we picked at around 500 watchers at youtube and 100 here in zoom so that's a total of 600. uh miss lara are we ready for the last part i think this is the last part right so miss lara uh nara's internet unfortunately got disconnected um so she's trying to join back in should should i uh all right leave her yeah so so uh do we still have any announcements or we had jp i'm trying to fish out some let's just wait for miss lara to [Laughter] okay can you try to i know uh the disabled officer um jp could you make her a host thank thank you for i know for saying okay so we're back to this um again what we have learned uh those exp uh design structures we have this crd and then our cbd so another basic design structure that we have to learn is this latin square design or we have this lsd okay so can you hear me yes yes ma'am okay you can see me too and then the slides yes okay okay i'll continue now now um just to relate the crd with lsd okay last time in crd we just randomly assigned the treatments to the different experimental units so we have this illustration here okay and then now for the lsd it's not just completely randomizing the treatments so there is this randomization again but in a systematic way so how do we do that okay okay next um in the next few slides i will be illustrating on how to systematically randomly assign the treatments to the experimental units okay just to have a picture it's a if you see treatment c here it should only be seen once in that row and once in that column right and then if you see a here that's that a must be okay once and that can be seen once in that row and then in that column okay so i hope you follow so it doesn't expect nothing about we have this lsd as the design structure okay uh recall again that what if i know that i i told you or i asked you about this what if this nuisance factor uh it's not just one factor so it could be two so that's why we have to deal with this uh we have to apply this latin square if you have two blocks or two um nuisance factors no so consider two gradients so those are the two factors or two blocking factors by using two two-dimensional blocking okay of experimental units okay used when experimentalliness can be grouped according to two sources of variation okay so if you can recall if we have an anova table aside from treatment error and the total okay so the source the source of variation you also have block right now for the latin square we have additional block okay so complicated anova so like rcbd blocks in the lsd must be complete blocks meaning in one row block okay there should be this presence of all treatment levels or in one column block we there should be this presence of four or complete treatment levels used for small number of treatments okay so the disadvantage here is we can only use this if we have small number of treatment levels why because this will also dictate your number of rows and your number of columns okay so the number of rows and the number of columns and the number of treatments must be the same okay so double parenting rose columns and treatments an experimental unit should belong to one of the classifications and to one of the columns you'll not we only see treatment a to that row one slang and then to that column once then so each treatment must be applied once each uh once to each row and once to each column okay so as you can see in demonstration here italian okay once for row one and then for column one okay application the same rcbd it's just that we consider another locking factor so it can be in the agriculture experiment biological or medical experiments engineering or any market research then we have this randomization so the same pattern of young discussion at them there's a c r d r c b and latin square we have this randomization then anova and then you have this test of hypothesis okay so obtain a basic t by t latin square plan okay so you should have this square plan and then randomize the assignment of row factor next is randomize the assignment of the column factor and then last this randomization of the assignment of treatment okay so how do we illustrate that so first okay we have this original layout okay so what i found randomization involved so we have this arrangement one two three four and then for the rows one two three four and then for the treatments abcd okay as you can see so this is the original layout and then what we do next is we get the okay we generate random numbers okay to have this arrangement for the rows for the row factor so if we generate random numbers the smallest number here is three seven five that has the so this um has this rank one and then next a second to the last value is this point six five four this has round two and then next round three and four so as you can see in the arrangement we have now one four one and four it will last around as a row four okay i hope not you can follow one four three two okay so that's for the row factor of course you will not use this when you use the star because they have this way not to randomize and then for the column we have this lowest value the generated value uh 0.231 and 0.089 that's the lowest value generated value so rank one okay so that's column two that's why here we have two and the next is two this is column four so we have two four and the next the three three and then the last rank column one so the interchange columns and also the rules no nakari nayeon saros okay so the next step is the randomization of the treatments okay so from the rows to columns now to the treatments how so we generate again random numbers and then the lowest value the generated random number is this one okay has rank one one one and then variety m3 so this is your treatment level three okay we have m3 here the next one next to rank one is round two so we have m2 of m1 so we have m1 here next to m3 and next we have m2 for this rank three and then lastly we have round uh four four and four this one so if we have three one two four for that column we should also have three one two four for that row for the first row can you follow so this is it now how about the rest of the cells in the table where where when we have this m2 here okay take note that the preceding treatment is m1 okay and m1 has rank 2. so what's next to rank 2 it's rank 3 and that is your treatment level 2 or the m2 so i hope it's not confusing you just follow the ranks okay and then what's next to rank 2 which rank round 3 is rank 4 so this is treatment level four or the m4 the next to em to rank four is go back to rank one okay and then we have this uh treatment level m3 okay so if we have this pattern two four three the same in the column in the second column you should also have the same pattern for the second row so two four three okay and then how about this so m4 okay so 2 4 and then what's next to m4 we have okay m4 it has rank 4 so we go now we now go back to rank 1 which is m3 so that's why next to m4 is m3 and then next is m1 go after rank 1 we have rank 2 which is m1 here okay i hope you can follow so this is now the way we randomize or randomly assign the treatments okay the experimental units okay so that's it so as you can see m3 can only be seen once in column of this in this column and in this row okay so that's it okay so we're done with the randomization so we just take note that if we have two treatments we have this original layout and that we have three treatments we have this original layout and if we have these four treatments which actually we tried not to illustrate we have this layout and then this is what i told you this um additional okay source of variation we have for the first block and for the second block okay before we only have treatment and error and then for the rcbd we have one row for the block and then now for the latin square we have two blocks so we have two rows okay so this is now we how this is now the structure for our anova table okay so the same pattern we now state the okay the hypothesis okay first we can use the muse or the mu1 equal to mu2 okay or the tau ice okay okay so for the ho there is no difference in the mean response among the treatments okay mean response among the treatment so you can denote this with this mu 1 equal to mu 2 equal to mu t and then for the ha at least one of the treatment levels has a different mean response okay and then test procedure again f test okay assuming that the assumptions are satisfied and then the f uh test we have this formula we have this ratio okay for the treatment okay take note the numerator here is for the treatment okay then reject h0 if the f uh computer is greater than the f tabulated or we can use the eval then aside from testing for the treatment means we also have to test for the differences um of this among this row factor so the block one okay so you can use this notation okay row i equal to zero for your ho so there is no difference in the mean response among the levels of the first block or the row classification and then h a at least one of the levels of the row classification has a different mean response okay at least once in the event from the rest and then each and then the test procedure we have f test okay take note that the numerator now is different from that testing for the means uh using the treatments we have this now for the row okay row factor we have msr and then for the decision rule reject h0 if the f computed is greater than the fta deleted or you can use that p value and then okay we have for the treatment our treatment levels we have the for the row and then now for the column so we can use gamma okay there is no difference in the mean response among levels but no difference gamma j is equal to zero versus h a at least okay one of the levels has a different mean okay so that's gamma j not equal to zero f test but then then we have this different ratio in numerator so for the ms column classification and then we reject h0 if the f is greater than the f tabulated or we can use the p band example assuming here that the assumptions assumptions are satisfied so we have this horizontality variances normality and independence of errors and this additivity okay and then for this four um applications okay they were compared okay in a variety of corn so in terms of yield sixteen plus were positioned four by four in experimental area okay so my four by four you can imagine the experiment disbelief that soil has texture okay so it could be on clay percentage and clay okay or until the unknown percentage you know lesser percentage of clay example and the moisture gradient so siguro young first so we have this moisture gradient and texture or soil texture gradient so this can be now your blocks that can affect the yield of this or the responses i mean and then this is used as the basis to incorporate two blocks again in your current experiment so we have this treatment row classification column classification experimental unit and the response variables so i think 19dhana matsukoro so in this example i assign the row at the texture of the soil texture gradients as row okay row factor and then the moisture gradient as the column factor okay then these are applied to plots kaya and plasma experimental units then the response is the yield now if we have this raw data okay as you can see my blocks okay for moisture okay a column column factor or moisture yeah you can see their columns one two three four then for the texture we have this row factor okay so every row so one two three four okay and then we have these treatments here okay t1 but i'm gonna summarize responses then we can see the row total the column total and then the total for each treatment okay now using star you can have this output here as you can see which okay source operation is now a significant significant result you can see only the row factor no that differs no in terms of the means okay so if we have this anova table again okay you can have the f or the p value not the design and then okay so we now have we can now use okay we can now test now for the differences among the treatments okay correction formulations because example okay so there is no difference in the mean yield among applications so that's your h2o no difference then each a at least one of the applications yielded the different mean yield result okay then you can use again the f satisfying assumption and then you have this okay so i showed this to you already and then the test is statistic we have 0.6 so this is actually the value here in the anova table so 0.6 and then the p value if you use the r you can out the star you can use the you can have this p value 0.639 and you can compare this to what to the value of the the set level of significance 0.05 so if 0.6 is greater than 0.05 therefore there is no enough evidence to say that at least one of the application has a different mean yield okay now among these three um applications or three treatment levels now next is the test for the row factor so if we consider row as our annoying soil texture gradient okay as our row factor so ho now is no difference among the texture level versus the ha at least one of the texture level has a different mean okay then okay and then you can you can go back to the star output and check okay the p-value there is 0.027 okay and so there is enough evidence to say that at least one of the uh texture levels has a different mean yield texture levels okay in terms of the means and then another is no difference in the yield among moisture levels so it'll column factor not in your moisture level so that's your ho then h a at least one it has or have a different mean yield okay and then f test numeration the numerator has this um c msc for the column factor and then we go back to the result using star we have this point six nine two as our p value and we compare it again to alpha point zero five and then there is no sufficient evidence to say that at least one of the moisture levels has a different mean yield okay so it's okay now at least in this example um treatment uh levels factor or some block differ so it's okay now to check the efficiency or the relative efficiency of this lsd over the crd okay we have this formula and then the notepad if it's greater than 100 then we can say that lsd is more efficient and if we have this value less than 100 percent then we have we can say that crd is more efficient okay and then okay so just to continue uh given this formula uh we have this anova output so we just plug in the values what's this so where is that point eight point zero zero eight the mean square okay the mean square for the column factor we have this msc okay and then the mse the 0.0016 this one here and then the msr okay the ms for the row factor we have 0.0099 okay and then we after plugging all the values we can get the result now which is 193.75 percent so that's way higher than 100 therefore mass efficient u lst so the computed relative efficiency is more than 100 thus the added variation of the blocks resulted to gain in in precision thus lsd is more efficient okay than the crd by about two times because it almost 200 okay so now i think we're done so this is now the summary first we uh we had this basic concept so the terms the principles and the the design structure so you were introduced with the different descent structures and different structures and then we also had this uh crd then rcbd then lsd and then final note okay statistics when properly used makes efficient research however it is not a remedy for poor conceived or badly executed research thus it is essential that a user of statistics clearly understands the techniques he employs then in case for my consultation um i would like to introduce this uh statistical consulting group which i am also a member okay so sometimes we have this free consultation we do this nationwide okay like last year so every anniversary or like every a special occasion of instead we have this after consultation but uh there's a possibility that there will be another free consultation or in case you can visit this website here on more details i think i'm done yeah yeah yeah thank you mama [Music] um i would just like to uh show a couple of slides about our finca research anniversary training program specifically we would like to show you all that uh how many how many are these jp seven to eight mandatory eight mandatory training courses line up specific to high school students and specific to heist uh to college uh students as well um just an fyi um and one more one very important aspect where we're trying to provide you with in all of these lectures is that in addition to um teaching you the traditional approaches to each of the steps in a scientific research project since um all of us in filipino science have are practicing scientists we are actually working in research and development um we also have university professors who are not only teaching but are also conduct actively conducting research um in the country are also like in other parts of the world we are also sharing with you practical tips and um not training courses uh we have covered ideation on this was actually back in september of 2020 so we covered um the ideation process and and and techniques uh at the high school and at the college level and um i i've been working as a research scientist abroad for about like 12 years now so even at the phd level is on how to research the literature so that is a philippines training course on how to actually use modern uh online resources relevant literature to your study aspects of designing an experiment or designing of a study um uh miss lara this is actually focus on statistics so what so in cases that your project requires that you have statistical design as a critical element um in in performing your studies the design of experiment introduction again and then our fourth lecture was very important um la lunas training program um as a youtube channel and then jp has actually done a great job of setting up a google classroom where you can take up all of these courses as well finish up the entire process you'll become part of the philosoph research university alumni group at dunpo um since we're also bringing in a community of filipino scientists and international experts from all over the world member puna amin research university uh [Music] community will have first hand access kona many more materials for a delivery number in the future that could help you develop your uh not only your passion but also your your expertise and comfort level when it comes to conducting research and then the last slide lanco so in addition to these training courses um every week we also deliver special topics courses because for us um for you to be able to come up with ideas um you know in in a number of different um stem areas areas we bring in experts from different parts of the world to talk about specific stem areas that you can specialize in so for example we had to talk on nanotechnology we brought in a filipina our renowned filipina expert when it comes to mycology and taxonomy uh last week we had professor gerard de moncast who talked about chemometrics we also had a very successful talk on data science so data science phrase uh was was uh delivered by professor kevin season and this is actually an up and coming um area in the philippines uh actually even at the high school level by the monster montreal and then as we said professor chad created next week cancer genome analysis we're bringing we're bringing all these special topics to you on a weekly basis so indeed okay we're gonna train you on how to practice a scientific method and apply it in research all right thank you for your jeff so i think we can proceed now with the q and a are you there okay um i will just give you some questions the first question that we have from youtube actually question is you know how how do we determine for the ideal and acceptance sample size in different research design so examples survey was experimental causal comparative and uh questions uh an adoptable acceptable number of samples um there are many cases it could be um although my formula for that given by your population size or given [Music] like for example if we're doing this um experimental design 3 replicates for each combination of treatments or for depending there is this there are many formula for that formula [Music] minimal number of replicates the value or the more the number of replicates or experimental units the higher depreciation so that would help us validate our experiment if we're doing the right experiment okay so so suburban sugar syndrome is so i i think you i like the the point that you made up i think that i think that law but then um yes it's about grouping the same or homogeneous experimental units it's just that's a field young soil [Music] is midland highland interchangeable talaga is the response variable synonymous with dependent variable i saw that question two oh yes yes yes is the sampling unit a subset of an experimental union yes yes or clarification specific question mom what if the result of fc is less than f yet p value is less than the alpha should i reject the null hypothesis um we can use the p-value anytime but there are researches or memory arguments about that or debate no now you cannot just dictate uh you cannot just simply compare that p value to alpha now point zero five point zero one point comma just okay depending on research for example medical research you have to be strict with your young mother depending on the type of research and the heterogeneity of your sample there is sufficient evidence or 0.01 sufficient evidence yes you have enlightened us so much with your lecture very grateful to you for helping us in our field and thank you from jos josue mirabite thank you for a very enlightening lecture hope you can also uh get an access of some of the materials [Music] earlier and then the good thing about that is that my mother played new york last question jpg consulting group now and then you can visit the website so you have that so that you have the idea of how much is there usually 250 line minimum a for one hour and then [Music] https uh uh instead that uplb.edu.ph extension services if you want to have [Music] [Music] because you pl andy and these are you people you people you you'll be professor you'll be statisticians so you know um i don't think you will ever doubt the quality i know undergraduate oh [Applause] you can always uh email her email the institute of statistics of uplb and trust me statistical analysis among private companies or private firms friendly public public domain why why did you use star instead of spss to analyze the data okay um first you i know libre star [Laughter] number two credible long star i mean librina credible case developed develop and they are from instead also yeah and and that's also like one very important aspect oh no we actually had a meeting with miss lara about this since i didn't pull audience i hello high school teachers and students college teachers and students so we're trying to make the content for as friendly to everyone as possible and it was developed by some of our trusted colleagues um in is it in repo institute of rice international rice research institute one last thing marty i think there's a challenge here for you sabine fields i have modules for statistics education with that we will be concluding the the q and a for you evaluation links are already posted in the zoom chat box and youtube comment section so please subscribe to our youtube channel and like our page in uh facebook and we would like to thank uh miss lara for her time valuable time and damning honestly consultations [Music] five hundred um i posted the link already for the evaluation so um [Music] and uh yeah i think we'll we'll see you next week for another special topic on cancer genome analysis all right so uh jeff any last i know you know um thanks for your time we're at magana lecture peaked at about 650 about 650 live participants museum you know that's how it is philippines we're trying to elevate your uh mastery of all these essential elements of performing research so um [Music] happy lunch [Music] [Music] you

Channel: FilSciHub TV

Views: 10,038

Keywords:

Id: OM8kvWyetpo

Length: 188min 10sec (11290 seconds)

Published: Fri Apr 16 2021