exploratory factor analysis in SPSS example 01

here comes an exploratory factor analysis video in SPSS okay also called the fa what I figure I'd show you is the vocabulary here in this preview to the movie here there's uh there's several new words that you need to memorize so you can connect your declarative knowledge with your procedural knowledge there's your college education for you but here's the vocabulary words that we're going to go over in this video in no particular order and here they are in alphabetical order so you could pause this if you want to implement but on with the show welcome to exploratory factor analysis we're going to put together this real quick video to show you how to run a an exploratory factor analysis in SPSS so we have a survey with 18 questions that are trying to measure depression okay so that's what the data looks like it's all ordinal let's take a look at the real data it just looks like that okay so and then we have roughly a huge sample size of about 2,500 of them above but let's go ahead and run this now first step is you're going to go to analyze dimension reduction factor all right descriptives you want the coefficient significance determinant and the Comeau and the Bartlett's extraction we want the scree plot first remember we're just doing a PCA now continue no rotation no scores no options now you got to get all your variables over there and do them all the same time so but everything's preset so we're ready to click OK okay so here is the output but before we look at the output we should probably look at the assumptions please hold the assumptions of a factor analysis our sample size and a general rule of thumb is anywhere between five to ten participants per variable which is per question okay so we have 18 question but we got over 2500 subjects so that's not going to be an issue with our analysis here reliable correlations we're going to look that up using the Bartlett's test of sphericity normality is not an issue here right it's not it's not an assumption of the FAA and the last thing is multicollinearity we will use the determinant of this matrix to see if it's multicollinearity or not but let's get back to the output first box this is your R matrix box or all the sent all the correlations between all of the items with two at a time so there's your sample size measure your km oh that is perfectly fine it's just got to be greater than 0.5 if it's less than 0.5 then you have a sample issue sample size issue there's your test of sphericity from Bartlett we want that to be significant and it is what this basically means is that you have at least one significant correlation between two of your items somewhere okay so that's a minimum numbers you got to have at least one correlation in there somewhere before you can run any FA so here's our commonalities communalities cim yux that's three word commonalities that's the right word box plus the extraction the strength of the extraction in other words this question one point four one eight that means that question one load it up whatever it's new factor was at about forty one point eight percent of the variance can be alluded to that first factor so this extraction point you're looking for any number that's less than 0.3 so if you have a number under the extraction table that's less than 0.3 you probably will have problems with that individual question or that I and I don't see anything less than 0.3 so I think we're good to go and here is our total variance explained so according to this table get back down there we should have one two three new factors those are the factors that have an eigenvalue of greater than one you'll notice here over once we hit the component for the eigenvalue value the eigenvalue drops down to less than point one so we looks like we're going to keep one two three but we're going to double check we're going to run a parallel analysis please hold we have this web address on the page in the middle that says parallel analysis so please use it okay so this first box is asking us for the number of variables that's the number of questions so we got 18 questions the second box is asking us for our sample size that our sample size is huge and we're going to click Submit over here and wait for responsibly so at the parallel analysis calculator page this column where it says means these are now eigen values so you're going to keep the first root the first factor the first component if this calculated I'm sorry if this eigen value is less than your calculated eigen value so we're going to keep the first factor because our calculating eigen value is 6.9 which is greater than one point 1 4 so there's factor 1 factor 2 we're going to keep it because 1.3 the calculate the calculated eigen value is greater than 1.1 so we're on our third factor shall we keep it or not so there there's is 1.09 ours is 1.2 so we're going to keep the third factor we're on a roll here will we keep number 4 yes or no there's is 1.07 ours is point nine seven six so stop we're not going to keep the fourth component in fact we're not going to keep anything past the third so according to the parallel analysis we should retain three new components or three new factors so hold on so now we know the correct number of new components to retain so we're going to redo this bad boy we're going to go back up to analyze dimension reduction fact or we move this over for you a little bit we're going to go to extraction we're going to put it on a fixed number of just three all right after we've picked the number of components that we're expecting we're going to go into the rotation part we're going to do the Oh blue mean again that's because there is a glitch in SPSS this is the only way we can figure out a way that would generate the table we need to look at and that is the factor correlation table or the component component correlation matrix table because that'll tell us whether our factors are going to be orthogonal or oblique so we have to pick the oblique which is Oh blue mean we're going to click continue okay you got to go to options we're going to sort them by size and we're going to suppress them we don't want any coefficient that's less than point four okay so that's going to be our cutoff for the coefficients we're going to click continue and I think we're OK to clicking okay all right so here is our new output part okay here we go so once we made those changes we're just simply going to scroll down to the bottom of the output sheet we're only looking for the rotated now I'm sorry the factor correlation or the component correlated matrix box to determine if our new factors are going to be orthogonal or oblique so we're looking at these correlations and my eyebrows are going way up there very they're not close to zero okay so they're not weak so they're they're moderate but again they're not greater than 0.5 they're close but they're no cigar so I'm going to go ahead and call this an orthogonal matrix again I'll say that again if these correlations here we're all all greater than 0.5 that means that your new factors are strongly correlated and you should stick with the oblique rotation but since they're not greater than 0.5 we're going to assume that they're orthogonal e related so we're going to redo this one last time and switch the rotation to let me move this over here what we're going to do is switch the rotation to a very max very Max is the orthogonal of choice and click OK and here's our new output let me shifty shifty so again here you get a lot of repeated measure a lot of repeated data from SPSS every time you go in and make any kind of changes so correlation table don't really need that we already got that we did that we did that scree plot we already have component matrix right how long they you know how they load up actually we don't even want that we want the rotated here we go this is our important box right here rotated component matrix so this is going to tell us which questioned is under which factor so according to this question one has underneath it I'm sorry factor one the new component factor one has question 516 11 6 12 8 13 so those questions are strongly related so whatever they're measuring must be relatively the same thing the second factor the second component has these questions underneath it 17 18 12 blah blah blah blah and the last factor only has three questions underneath it 7 9 and 15 ok so we're going to go the next step you but I want to write these down somewhere which questions which items load up under which of the factors because we're going to use that data to run a reliability test on each of our factor so give me a second and we'll get that going all right this is going to go fast checking the reliability of our new factors you're going to go to analyze you're going to go to scale reliability okay we're going to pick the questions that load it up on the first factor factor number one and co-pilot please read them to me 5 6 8 11 12 13 16 and that's it you go to your statistics button the important one right here is this scale if item deleted in other words if you don't reach the chromebox alpha that's your that's your measuring unit we want a minimum of 0.7 of on chromebox alpha if you're chrome back self is less than 0.7 this scale if item deleted becomes a very important tool it will tell you if your chrome backs alpha will go up or down if you throw out a specific question so you don't have to worry about that unless you have a low chrome Beck's alpha but let's see what we can do here just click the continue button click the ok' button and it should spit up a chrome backs off a box and there it is right there BAM so the good night the good news is that our first factor has a very strong cronbach's alpha which means it's highly reliable and are seven questions that go into that first factor so that was good for the first one let's do the second one we're going to go to analyze scale reliability we're going to reset this and I'm going to pause this while I put all the second set of questions in there please hold well it wasn't as bad as I thought okay statistics scale if item deleted click continue okay so our next factor factor number two our Chrome back selfi is plenty big so it's point it's greater than 0.7 so that's all we need to know and it's got three questions underneath it looking good looking sharp so we're going to check the last one now we're going to go to scale reliability we're going to reset our data and through the magic of technology I'm going to enter all the questions that load up under factor number three and a blink of an eye amazing amazing so these are the questions that load up under the third factor statistics scale if the lien it could continue okay all right let's check out that crow oh we got a problem Houston chrome backed off is only 0.6 we have a problem we want it to be at least 0.7 so what that's when that item if deleted box comes out that's the one right underneath it so let me pull this up a little bit so according to this wonderful little box right there if we got rid of question 2 our chrome backs alpha for the new factor number 3 would jump up to 800 and that's what we were wanting so that's what we would do we would delete question number 2 from the survey so in order to make this a more perfect survey we would delete question 2 and drop down to 17 questions so in a nutshell what we have found out from this 18 question survey that's trying to measure depression that it looks like it's it's measuring three unique dimensions of depression such as maybe work-related depression relationship related depression or maybe financial relationship depression so a very important last point here is you as the researcher you got to go back and look at those questions that are under each of the new factor and you have to read each question and try to decide what is exactly the common characteristic that each one of these questions is trying to measure okay so after you read the questions you should have what we call an umbrella term or something whatever that that specific thing is measuring so that's again that's up to you the researcher but that's it for now wait mgz and copilot out of here say something all right okay all right we're through by

Channel: Math Guy Zero

Views: 152,850

Rating: 4.8402205 out of 5

Keywords: exploratory, factor, analysis, SPSS, example

Id: 16Fbz65rTck

Length: 15min 14sec (914 seconds)

Published: Thu Mar 26 2015