Tidy Tuesday live screencast: Analyzing caribou locations in R

hi I'm Dave Robinson and welcome to another one of my screencast where I'll be using our in our studio to analyze data I've never seen before as usual the data comes from the tiny Tuesday project an amazing project run by the R for data science online learning community and this week we're going to be analyzing caribou locations I'll first say I don't I know very little about caribou are caribou similar to move at a moose are they may be so it looks like it looks like they're mostly in Canada pink so this is a caribou and we're gonna be find out where hair bows are located and I'm really excited about that let's see okay individuals and locations all right I'm going to read it in library tidy tidy first and game set game light steps I like to start with and bring into tidy Tuesday data looks like we have two sets of data we're looking at today not to mention that as usual this is live so if you're watching it live I really encourage you to join the live chat bringing some bringing some comments and I definitely ask questions as you go I won't be able to answer all the questions but I try to answer questions to go live you've ideas for analyses ideas for visualizations please bring those in it's part of the fun of doing a live screencast so let's say the two datasets that we have one is of individuals looks like one for each animal okay we have death cause of death do we have when they were fitted with the location tracking tag oh that's really interesting we're gonna have actual like combinations of longitude and latitude where they released comments or when they're deployed and then off oh these really cool gonna be like an animal tracking data set where we see like a hair group starts here where how it keeps moving how it moves over time that's really interesting I might do something like an animation about it I hope to see and there's like types of deploy off could be have a mane the animals captured up and find ended with death tag stopped working fell off alright anything other person removed etc okay we then we have a dataset of individuals and locations okay so let's take a quick look at and locations is going to be like so we have a timestamp a longitude of latitude and yeah I made a video I did to fire the links back to that first individual as well as it looks like location it's a hotbed to me that the study site is in the locations rather than in the individual but we'll take a quick look at that the main really exciting thing here we have is timestamp longitude latitude latitude I don't think without a dataset quite like this one entire Tuesday so I'm definitely excited I hit the close we have is volcano eruptions we have tough time volcano erupting and there I definitely did some animation but you know what I'm gonna do is I'm gonna start by exploring the individual it's really good to learn what we're working with here 286 caribou well they I've seen right away is we have some missing data in fact all right the vast majority are missing I think the best probably missing sex life stage ok well let's actually check how much missing data we have in these columns it'll help get a sense of what we know about them so what we'll say is there's actually a great function for this summarized app what is it it from oh right it's across Oh yep so the new version of um let's let's remind ourselves how does a cross work I actually don't know that I've used it yet it's cross so let's see an example summarize across this and then a list okay oh all right let's try this summarize across and then okay someone's across sex to study site let's ask a question let's ask mean not isn't a dot how many are not missing back didn't work it didn't come even a little bit close to working if I'd said summarize list mean it's like a cross didn't uh-huh hold on cross oh whoops cross this is the first argument aha I mean he would go can't compute how who says across this oh there it is okay what I need to say is like not yeah so we're going to get like not missing and then name is where actually I got like one right now we can actually say it looks like an customize the name so saved our names is I just wanted me to call them odd to me that is one two three actually don't quite know why why that's the case okay but yeah the story is there it's across these I want to apply this function oh it's pretty that more or less works so it looks like here we can actually show is that very little the data have pregnant and I wonder actually count pregnant it's like three folks is sixteen true and a lot missing I don't know whether it's soon the missing is true of feeling me we we kind of like not much what I do there it looks like and all of these life stage mostly missing okay so it's like we really only haven't got a quarter of the data for these with have really have a quarter of the data for these and then death cause alright so maybe I would definitely well death cause it makes sense it's missing because we don't know the reasons that that it stopped I can actually get back to my get back to my cruise data and look at our our documentation this is really happy the UM the tiger Tuesday package deploy off type okay so I want to see what is the reason these are the rituals it looks like mostly unknown summer removed sixty died a curious of the ones who died do we have that the death cause if I say deploy off type equals dead do I have the death type that cops often not always okay that's still pretty good a lot of time it's unknown suspected predator etc couple are great one killed by a grizzly bear okay okay so and of the reasons that they generally were removed it was like they were unknown another not very helpful and about a third removed about a little less than a third would because they died okay that's something some information we have about individuals we can see there's enough missing individuals there mom I'm gonna make any graphs I'm not gonna make graphs except that I am gonna ask about the individuals I want to ask study site I want start looking at the location data and go so no yep so it's generally like here's a quick head Graham I'm curious for here did they have do they have multiple longitude employee on latitude do they have multiple for you just study site and the answer is um largely yes it looks like but close to each other so I can actually I can actually plot this and say let's plot employ on to do longitude is the X and deploy on the latitude as to why size and she owns point I'm not actually Matt making a map yet this is this looks like a bug I highly doubt they went down to who knows what this is this is Texas or I don't know this is below 30 degree I don't know well I think maybe sexual maybe Central America I'm not sure this point looks like a like a bug but the general stories by the end if I say filter deploy on latitude greater than 40 we kind of see the shape general sort of shape of Canada maybe and yeah and we had borders that in a later at visualization but there oh I'm the other thing that I wanted to do is say color is the study site yeah so this is actually pretty helpful I'm going to I think I'm going to add I'm going to be add and turn this into a map is that we see where's it seven one two three yes seven study sites and they're generally like co-located so it's like these were all studied by the Kennedy site these were all by the quintet site another site is one specific location and then the people study and go out from there but could be something along those lines okay that's uh that's I think interesting interesting enough there let's start making it a map theme map I do I've done this and sometimes it bites me to do theme map I'm also gonna actually quickly say scale size continuously guide equals false I'm going to move the the legend for the size and if it's anything this is showing locations but it's not it's not a map yet so I need to add a I need add borders and I can remind myself really quickly how borders works with a region there it is database world regions Canada I think is going to be the story ok so here's what here's what I'm doing we don't say borders world regions I'm not sure that needs to be capitalized or not oh that's that's great and uh one thing we see is that all of these are an incredibly small region of Canada actually so small that we understand of it that outline or have that Allies in the u.s. I don't like this outlier I'm gonna keep this outlier out I'm really bad at geography ok so one thing we learned is in his other scale of Canada these are and are really small in a really small region see what region are they in there was probably in the documentation and I skipped it no it's not a documentation she was data has the documentation one set watch are not may see so let's look for a moment on this someone's Justin it might be Alberta I'm again I'm quite bad at geography so this is just British Columbia Ministry of Environment and climate change all right so if we're doing a map so this is like like tons of caribou and then for doing a map prolly one went just on some local region of British Columbia I wonder I wonder what map data there there is oh is there yeah all right so yeah someone suggesting province British Columbia is that a thing we can do is say border so normally I would say world regions is Canada but if I said province is that we love States nope let's actually ask to math of max package Canada so these are things in the arm here though Canada no but I'm cable Canada another perk a dataset of cannabis cities I'm looking for I have US counties I have a map of France Clements privates province there's a way to grab additional maps I also don't know like if it's going to be much for how interesting it's going to be shaped file yeah me baby do this this shows how we would go about adding a shape file alright I think let's see I think if I bring down the whole one some way if someone can remind you oh I've done this you know how did you stay before it what if I do let me see genome SF I'm gonna look up find one where I've done that done it before ah look at me I thought oh people did ever see that before in our studio command shift F on a Mac finds and all the files of frost this this our studio projects which for me is data screencasts all the screencast I do retire Tuesday and then I can actually dive into them so if I go in here Central Park SF kms up uh-huh yes so if I do CD downloads of CD downloads unzip what was it called Anna de Provence cool and if I then take this line haven't saved yet caribou locations if I think this line and I say this is can be useful in amis a province shape file is it a set with a tidy SF or what is that let's find out does anyone know I think it's library I sell Oh No oh huh downloads oh I baited a whole bunch of files make their province move Robyn stop everything in the province LS province that looks right aha province shape file now yes took only 47 steps for me to figure that out yes yeah that was that was about me by the way not about the craters the sf packages are really really handy these are really handy package in a really concise approach because I can now say GOM sf data equals province SF well I can't quite do that I can do because the problem is that it doesn't have this this this I'm gonna try put that in the G on point and cross our fingers just how big is this province shapefile anywhere how many points are in it can I just crash our I might have that's me trying to fix up I made a crash daughter that happens happens to me a lot because I don't know what I'm doing yeah okay I think it might have been because I resized it at that moment how many objects are in here okay - which is - Alberta and British Columbia - features 7 fields and it has geometry I wonder I'm just trying to like I'd like to see well what is the geography for each of these ok what does Jim SF take Jim I said I really don't know my way around these in cases in case it's not abundantly clear uh-huh uh-huh ok geometry I see ok yeah so it does um if I want to do just British Columbia I think I can do it believe I can do province SF filter for name equals EC and now I've got just look the bed didn't work what was province esse what was the name was British Columbia not Lisa oh I three said frog because BC and now I have the one one object okay I'm going to try that kind of save in case this crashes and try this again why is it not visualizing I was really excited oh we started doing it I'm gonna let it sit for a second I was very excited about this I think it might have like some like too many points you know that doesn't look how I wanted to even though I put the points on where are the points I don't see the points also it's got so much detail which is really about traffic but make this just British Columbia what a remarkable not a detail but I don't see my points on top that's that's back to Canada hmm maybe I should have dropped this and then I try this again wait for it wait for it this is too much by the way this is like ah and I don't know how to subsample from within the job dude geometry that part I don't quite know waiting for it yeah I'm not getting success hearing like I've seen this little point and then where's on my actual genome points hmm alright I'm gonna drop these I said I didn't hurt the individuals finesse a fabric that shouldn't ah so sorry I've heard a couple of couple suggestions one is are they under the province shape I don't think so because I put the points second it should have been put on top by usual TV at least that's where you did you talk to you we usually work is I put the point afterward so I said you convert the individuals for an NSF object I don't think I have to I should be able to in the squirrels I plotted both the points and the SF on the same graph on the same graph TMS app Jim point so I don't have a I don't know why this happened I am gonna go ahead and skip that and and we'll just work from here as like here's a longitude and latitude in fact don't drop the theme math though it will say you know I'll have beam light I will say theme even the panel is kind of away so this is this is like kind of a kind of reasonable we were looking at here's the deploy the deployments all right I'm gonna stick to this right now all right sorry to get sidetracked on the on the map beam hi that was looking at individuals that anything else I want to know explored realized by individuals we found a protective pull off type did you check the study I don't see Oh animal ID hopefully not oh okay we have duplicates of some animals and let's find out what this individuals filter animal ID all right this could be the same animal because in all three its female it goes from like Stage three to five presumably this year's study site is the same the locations are similar and we don't have the time the deploy off time see and that's actually a bit of a frustrating thing is is if I'm going to it looks like yes it looks like recovered field collar replaced the new one it definitely looks like these three are the same caribou but that's gonna be a problem and we join with the other table because I'm actually realizing we go yeah is let's take a look at view what is it called location locations its that location its individuals I guess locations this is gonna be a problem we joined this data together I just want to point that out in advance is that we have an animal ID and event ID but we're gonna have duplicate animal IDs if we try joining them back okay so that means that for now not actually going to look at the individuals I'm going to start working with the locations what I'm gonna do is take the locations and visualize them on some almost exactly the same way longitude latitude heck I'm even gonna do color equals [Music] what was the study site I kind of liked how was riccitellis regional there and genome-wide 250,000 points oh well hi this is so cool I did not realize that it was this densely packed it looks like it's about every four hours I'm not sure how long the young the ranges are but that's really cool ooh somebody gave me some code when one more look well look at this I'm gonna die I'm gonna jump back into this but somebody gave me some this is um why cease oh I see say some so sorry I'm a potential name well did it gave me a bit a code on us okay I had to try this once because I am I'm interested in the but but here's my question for this this comic here have you tried combining that with the GMs F in Provence and they like the provinces that's the step that might not work okay this is one thing you try it's imaginary numbers on Twitter is the is the helpful comment job and then come back and try that but I'm good now I mean I'm going to keep on this path because we saw the province does is pretty large relative to this okay so the UM things works for that help yeah okay so let's take a look at oh yeah looking within each of these regions we have like you really want to know changes over time and I might want to pick one individual so to do that I'm actually gonna say take our location I'm actually curious about something but group by individual and why not what individual actually called animal ID and let's say goodbye study site when did they start being tracked when do they stop being tracked that will be an in timestamp max time step and for that matter numb points is n so when do they wind and I'll unglued something is slow than this so then it should be it's the video buffering let make sure I'm on the right one moment I am right next to my router so I'm hoping it'll be a bit better is to tell me if the video is doing any better Oh No and it has to do it how's it doing now can people hear me he will hear me great okay it looks like I'm back the last thing I do is gonna put in the comments I'm going to add the link to the shapefile I decided to link to the to the shape file they didn't want to follow along with my code alright so thanks for your patience and do tell me yet tell me if I ever end any other problems okay that's right actually gonna do is filter for animal ID is let's say one sample animal ID one this is actually really easy way to say pick one single caribou at a time technically if I don't want to be weighted towards the more frequent ones I do sample unique one so then I would have to do is say okay let's let's look at that one caribou over time there's actually this I think a really nice thing to do is a longitude latitude chiona there's actually an interesting trip if you want to show the path that that animal takes use genome underscore pad not Chi online why because line would look would always go up and down and up and down that's not accurate because this person this moose is not traveling only a one direction on the x-axis instead used to you on behalf to make absolutely sure of that we could arrange by timestamp first yeah so um I can actually say genome point and G own path oh wow and what's fun about that is it kind of like oh and I think she say let me see Gian point she um path watch this I'm gonna actually say so example and I'm going to keep the animal the same as we as a mess with the visualization so anyways take a look at this path and then I say colored is the timestamp and this will get a gradient of like from the start to the end that gradient isn't your cup of tea we can try a scale color I kind of like grading in two low equals blue I think maybe my default lowest ruin high as rabbits find out scale color gradient ooh let's make it as America time stamp and let's say medium time is median of example animal time stamp yes no I don't like any of this anymore because what I think I have to do is here we go [Music] promises where they want to be visualized know that that gradient is just not gonna work skip that yep totally step back never mind this is fine maybe we'll try something else in in a bit but yeah okay so this is one it this is one animal and one thought here is like yeah I'm swimming around it a little bit like that's pretty that's actually this is a pretty cool world you can see it like kind of varying I wonder if I tried but if I tried doing actually a year of time staff hello hi can people hear me now hello I think I'm back I think I'm back let's see hello can anyone hear me and a way to tell someone someone tells me that they can hear me in the chat otherwise I don't know if I'm if I'm actually going going through and I don't hear me hello this is gonna look great when people watch it later just uh it's gonna look phenomenal you didn't Malcolm into the chat Oh people can tell him me they can hear me oh okay chat wasn't loading alright fantastic they actually believe it or not I don't think it's happened before that I had that big a fiasco thank you for everyone's patience those of you who stuck wit through it through with me okay alright so the plot does plan on making it was here we're actually showed this is like trends over time by month is not is no good let me try order which is kind of by season aha this is pretty good okay I can see the chat now thank you so much sorry about that oh wow alright so the UM again figure everyone for telling me to do that you can hear me I was kind of his where Devin was sitting there little having lots of tension well alright so here's the visualization I was looking for it this is one moose over one part of me caribou over time when you say title is one caribou over time yes and then what's interesting interesting about this yes we can see that like for where's red it's hard to see where red stuffs off it's a bit where the cable starts out here and then kind of goes it ends here is kind of a story that we can see and yeah this visualization is kind of solid not perfect I think I think one of the issues it's hard to tell the actual directionality but I think it's not gonna be any easier if I if I try making it if I don't make it a factor it's not that much easier to say like here you get more of a sense of the directionality but you get lesser sense which I could have done without the floor date and yeah the this would be looking at it over time okay all right so that's um that's an examination yes one cab over time and here's fun we can actually try a different care of over we run this entire thing and we look at oh there's another caribou that was entirely within one year I'm just gonna drop this yeah my try would continue with a continuous one there's some that span years this is pretty cool how it like actually shows in two different clusters someone said what if we try alpha-a instead of coloured one thing I think is that's Jake one thing I think is that if we do alpha it's gonna again add both alpha and color that's not gonna be go from transparent to strong I think it's a little bit it's kind of hard to really kind of see here but that doesn't actually give me the idea that if you make it all the points with alpha equals alpha it represents transparency if you make them all transparent that gives us a little bit better of a sense of like what has just one point and let us look through the the various points I think that might be a reasonable thing to do but I'll have to do that with path too all right so that was a little bit of an analysis and so we'll just what if we fastened wrapped by quarter we could my fears if we do let's sit so I would maybe faster that by year instead of a quarter I think by my corner is gonna be a lot but I can actually say like well actually in this one I wouldn't be it would be so much you know obviously I'm thing of a few ways we can try visualizing this but what I try is guess we could do whatever you floor date in Florida Dave it's good to use a stop date around and to make it a date object time stamp quarter and actually don't know that that has that date matters here I don't really remember that mm-hmm this step worked let's see it does not like what is it doesn't it like I wonder what if I did try a skate oh I was right at the first I was right the first time it doesn't like fastening by time step but it does like fasting in this way oh that's not that's not that's kind of interesting oh oh I like this so I think the big thread suggestion that's a bit quarter it might not work there's a lot of data but in these so far it is kind of interesting where it's like in age three month period where do they spend their time that's pretty cool okay so this is um this is some visualize it so here we see like I actually think this care was pretty cool and it actually gives us anything an interesting place to look at next which is to look at how much are they how much was the caribou movie total distance traveled I'm curious this data is it always every four hours I really that it to examine some cameras and start like looking looking at them because I think I'm curious what is the gap if I say arranged by timestamp what I'd say then is filter animal ID is actually let's I'll just use my one example animal now that think about it and I'll say gap is diff time of timestamp with lag time step here's a good trick unit equals hours count gap all right so it's usually oh I was wrong with exams for sometimes there's four okay this looks like four but this looks like it was mostly around as numeric this is one animal the gaps or imbalance biggest scale let's love ten usually one number but some occasionally rarely other numbers I'm actually going to say it going to say I'm gonna round it and count that so always gonna be an n/a for the first value or not is any gap why not it's true it looks like I for this for this caribou it's usually seven our windows what if I tried it on all animals what I did is I say I throw in a group by animal ID and I neglected to ungroup after the mutate alright so on then I did filtered act less then gap and Beyond call I know I don't actually need a count at this point yeah I don't need a can't count I need you want mr. Graham and this didn't really need to round it either okay so the thing with equals two okay the story is most of our gaps are we're tracking our caribou in like four to eight hours something like that but this gives me in general and idea I'm gonna say locations with gaps I wanted to say is take our locations group by animal ID you paid here we have we need a gap in time and I need a distance this is one where technically to do distance on longitude and latitude so I gotta actually say last longitude equals lag longitude last latitude equals lag latitude I'm gonna do Euclidean distance which is not quite real because because they are but this on these these tend to be very very close so I think it's gonna be really it's gonna be pretty close if I say longitude - promise um distance oh yeah there's a package for this let's see if I can remember it not they're not lost what's wrong I'm not gonna be doing I'm gonna do distances if they're Euclidean and it's going to be it's going to be imperfect last longitude squared plus latitude - last latitude squared square root of that old then it's not really done in speed the speed would then be the distance over the gap this is not in miles not really in any unit it's in and gap as numeric I'll say hours don't need a round sleeve like that me try to remember this tell me if there's buffering I think there might be a little bit of buffering what we're going to see and if I say let's see a lot of you the latitude distance R remember there's a there is a thing for this geosphere oh yeah that sound that actually does sound it absolutely sounds right fear I remember using the fuzzy join package and then if I say distance of it of a point green circle shortest difference between two points an ellipsoid but I don't need I don't need I don't need highly accurate I want to be as fast as possible it's going to do it quite a bit ah this is this feels pretty good have a seem okay can be a matrix of two columns whoa okay so we figured this out all right sorry that I got slow for a second but um yeah what I'm gonna do actually is what I was gonna say is I'm gonna do distances but I'm gonna do that with the disc the disc to you a disc ever seen I'm gonna try a bassoon and yeah I'm gonna try I say vaguely remember that being a good way to do distance know what I'll do is I'll say library geosphere I'll say dis haversine I'm Dulce bind and it'll be a matrix of longitude latitude see bind creates make last longitude I'm gonna call it last lon last lap just so I don't get really last long last left and it compared returns that I believe in miles so the Ono meters let's go meters this is like m/h meters per hour and senators are climate change into km/h it's in Canada's only fair use kilometers and put this over a you're gonna meet it I'm gonna say km is this divided by a thousand and say kph that was ridiculously fast is that really true I'm also going to want to only look at the ones on group filter hours is let on we'll see ya what are the speeds aha here we go how fast do our how fast do our animal to our caribou travel so this is how fast to a caribou travel and we say hello first title caribou travel hell on average halfa on average so another thing is like this is including sometimes they'll be sleeping somewheres don't walk one way then welcome walk the other way so when they I'm actually gonna do with that is I'm going to filter for a gap not including there's multi-day gaps those don't feel like they're a gap is less than let's say less than eight it's called hours yeah I think it does the reasonable start alright well you see is there speed on this scale is roughly log normal scales equals I don't like I don't like the um this format does this work this doesn't this oh yeah dad that looks better point zero zero once like the the average caribou travels a little under a tenth of a kilometer per hour it's just not moving not moving a lot that's that's Annie this is for every single measurement this is not the average this is not advocated across caribou but it could do now is take the locations with gaps by animal ID and summarize the same way earlier did our examples have I can throw in a study site but also add an average speed mean kph and nao they're all going to have at least one and oh I forgot to clear the filter gap it is less there you go to AIDS yes now I'm just gonna I'm not gonna do that this one yes I am I need to be hours less than babe except that I only want this on hours as a date here we go all right I only want this as a bracket as like a subset of our km/h okay so then I'd say here are by animal with a grid of the animals and now I can look at what are the fastest and the slowest care go and so we're going to do is look at average speed as a distribution it's good idea the high ones are probably ones relatively few observations that I can't be sure so I'm actually going to say numb with speed is some of ours less than eight one thing here is I only want to look at the ones who I have this this window of time and I'm actually gonna I am going to simplify it I am just gonna say filter hours listed with AIDS we know all the ones what our gaps don't dumb don't line up and why we doing that because I want to say plot the number of points by the average speed not a histogram if you have a point and a scale x log 10 so when we see here is we definitely want to include any of these these are once we did we didn't have out of like eight our tracking data on so maybe we want to say at least definitely least like 10 filter non-points so this is better and this shows our cloud above our faster and our slower than faster-moving caribou and then we have two that are really higher the um and I'm really curious about those two especially this one it has a lot of points but it seems to be what is this incredibly fast caribou arranged ascending average speed and we have is yep we have one with 89 and one with 2684 that's the way I'm worried about this one we track for three years have a lot of data on it and is it's just a really fast care but let's find out or or as though or possibly is there one point that is say in Hawaii hit my accident that makes us think that it's a far greater distance well we can find out with location with service locations with gaps filter animal ID is this and we're stars like an arranged by date by kph and view and see did we have say a few extreme points we definitely have some extreme points but there are there's more than one and none of them are completely insane so the kilometers are likened at point three two yeah how did this but how did this data when it looks like kph quickly goes that not only 0.40 what is the gap how many hours are there uh okay a problem with this data check this out notice there are hours look like they're wrong where we actually have a gap in time between two points that is only a small fraction of a second and this definitely looks sketchy because there's no way that these points were right next to each other and yet the points are very distant that sounds like that like it definitely sounds above so that is a data cleaning step that I'm actually gonna say let's not even trust two points that are too close to each other in our analysis say ours must be at least a half an hour apart half an hour's our start if I could say at least an hour but it's very least it'll fix that one and it does fix our graph we now no longer have that point being an extreme outlier we move the cases where it was like two points the right mix which other looking like it was sitting around at the speed of a car all right so the UM I'm gonna add one more thing to here expand limits I'm not gonna do x equals zero because oh no you can't do execute on a lot of scale but yeah this is speed of care group and it looks like yeah there are some faster caribou than others so then I can pick say a fast caribou but we say by animal arrange an O and I'm actually gonna throw this filter up into our aggregation that's a nice step because now I can say here by animal look through here num points greater than ten and I say by animal filter and so I arranged ascending km/h Oh a bird seed and I can graph what does this incredibly what is it's not incredibly fast but I'm usually fast caribou look like I'd do that with one of the same kinds of graphs that we do other places and that graph looks like here we go by animal locations with gaps filter qu car 107 this is a fast terrible here's something fun I'm right now coloring by thereby oh yeah looks like there's some jumps here I'm coloring by the timestamp what if I colored by the speed I should have done kph yes this actually tells us something which is that most of this Caribous um time was spent not going very fast and there were a few really kind of this looks like one extremely extreme value I obviously think that's kind of what this suggests I'm not 100% sure and say ggplot kph filter animal ID is qu car 107 scalex log 10 yeah there were a few extreme values here hmm okay that's worth worth noting but the median is still pretty high okay so this is still is a fast care though alright and my other questions are gonna have is to care about travel at different speed during different seasons this suggests that they do like not necessarily I'll do but notice like oh relatively slow relatively then here moving around more than individual clusters deceptively state cluster sections where they move around so the next thing that I'm going to do someone else just a bit colored by day nighttime I think that's going to have too many alternations or they'll be literally hundreds of alternations with them or at least a dozen many dozens of alterations within each of these facets so by the way one thing I notice I really appreciate some people have been have been providing code which I really appreciate you share with me like thanks buddy but I'm I just haven't had a chance that we're gonna be the analysis since I've been a bit short on time due to the technical difficulties so oh yeah the last thing that I'm gonna do is I'm really curious but look at this if I group by the month of the year does are did those typicals of the distribution of typical speeds change I'm gonna filter for only the reasonable subsection of hours and then I'm gonna say group by month is month of time step that's month on a one-day trial basis I'm gonna summarize average speed is mean of kph aha there is we're over a year by the way does month is month come up get labels yes and group equals one so here we go January through December there is a trend where it looks like the animals are start slow and then get and our fats are slow in the winter and the spring they're faster in the late summer and the winter goes back down December then back down to January since there does seem to be a seasonal trend in caribou speed so that's really cool there were there's just look like caribou speeds by season and caribou name I don't think we have the names okay we have um species in here Phillip back individuals we see yeah I don't think we have species here but I do think we have we do have a study be let's call it the UM study site I'm gonna add that to this group Ian's study site and then I'm going to say I'm gonna get that and get the also that the end it's always good to look at at the end as well and say giome here we go Group II go study site and you know point size equals animal color I've got color equal study site and good because study site that's necessary fertilize so it does look like there's a it's a good thing that I've looked at the study site for the UM it's a good thing that it looked at study site because we actually do notice that some the travel faster than others I could have done a facet wrap maybe it maybe drop the legend so it does look like some of them were did see there were seasonal trends in caribou speed as well as differences in the in the study side I'm gonna tell you this is a pretty once where's the average speed traumas per hour on the elbow y axis and put the labs at the end okay so I just wanted yeah I want to show that it's one aggregation there's a lot that we didn't do here I didn't look at him I didn't like cluster it and say oh it's likely that I'm that care who spend time in one place then move allowed them to town a place I didn't make an animation which we were which we were definitely talking about and and there was a package suggested unfortunately I said it was a little short I already went a little bit over but um I was short on time because of the the technical challenges and the what was the name of the package some people are suggesting a package for visualizing an animal movement so we wonder what behind me it was something like animal or something like that it looked really cool and I would know what the package was I'm looking through a cat Cinda a move Vince it was called yes so there's a package move this that I didn't um try out and that definitely would be a it's really cool look at that you make little squiggles watching it oh that looks like so much fun I don't have time for it I think we should have won this already to stay I've been in the States they know less thing along here so this is definitely something worth trying when you can see different kinds of values I've bet I can do things like the speed over time that's really it's really fun I've never worked with them as you can probably tell these are very simple analysis I haven't worked with things like animal position over time and yes this is just a kind of some starting points all right so what we did is we visualized in an individual individual we were just starting sights we saw they were regional we looked at a single animal that we randomly sampled and looked at their positions over positions over time we saw that in some seasons we actually see can confirm here notice that the notice that in the first six months the year this caribou stayed in the same place and they need a lot of migration in the second half of the year and that's true and yeah it's a little bit short each of these there's a lot more that we can do there's a lot more that we know we could have done in there my last thought is I wonder if the median speed if medium speed would have been giving me a different story that's weird uh yeah no I'm gonna stick to Apple you know six average but no I like medium I'll tell you why I like median is it just remembered yes don't use average speed because we saw earlier that the speed is log more log normally distributed so that means rqm too extreme this is a weird this is a weird why I kind of want to just drop that one I bet I don't know how many points were there but I am gonna say heck I'm just gonna say filter study site is not equal to heart they just thank you for your patience while I finish this up pardon well were there eight sites yeah so this is a story this is just kind of a little story to tell all right so this yeah and this looks at seasonal trends that they looks like a different across different sites it's hard to say whether it's they got they found anything type of caribou different regions they've been something else but it's definitely a factor that seems to have it seems to influence it many of them have a dip in June don't almost all of them except for Graham don't know if that's real or not or possible there are some kind of data issue and yeah we created a couple visualizations no animations and only the simplest of maps but this was so really fun like looking at animal tracking data like I said is something that's pretty new to me I could there's so much that we've done here alright that concludes it thank you for bearing with me through the technical details I hope you had fun I certainly did I'll see you next week

Channel: David Robinson

Views: 2,396

Rating: 4.8769231 out of 5

Keywords:

Id: 7G7SVODhVo4

Length: 62min 34sec (3754 seconds)

Published: Tue Jun 23 2020