Hi Dagmar, I did not understand your complicated sub-sampling, sorry, but I would suggest check other things first.
You mentioned 2 animals with 5-minute fixes, total 10000 fixes (about 34 days?) and other animals with 1-hour fixes, total 150 fixes (about 6 days?). Maybe I did not understand your message correctly? However, not surprising to find animals tracked for longer duration show a larger range of movement than animals tracked for a short duration. Did you track all animals at the same time of year? If not, your comparison could be confounded by seasonal changes in activity. What species are you tracking? If you have fast animals with irregular movements, then 5-minute sampling could capture excursions that get missed by 1-hour sampling. My calculation of duration (above) was based on consistent fix success but that's often not true in reality. Tracking devices sometimes do not receive satellite signals at certain places/times of day etc, so it's important to check for missing fixes. If some animals had more missing fixes or different patterns of missing fixes, that could also confound your comparison. If all your animals were tracked in the same season, for the same duration, with consistent fix success, then it would be a valid strategy to sub-sample the 5-minute data at 1-hour intervals, rather than random sampling. With one-hour sub-sampling you get the same representation of behaviour for the "5-minute" animals that you would have obtained IF you had programmed their devices for 1-hour intervals like the other animals. Best wishes for your study. Julia ________________________________________ From: AniMov [[email protected]] on behalf of Dagmar [[email protected]] Sent: Tuesday, 29 August 2017 19:03 To: Mathieu Basille; Animal Movement; R-sig-eco Subject: Re: [AniMov] [AniMov/R-sig-eco] adehabitat HR: Problem with BRB and sample size (number of fixes) Dear Mathieu, Thanks a lot for your answer!!! I did something similar like you suggested: I subsampled my data and tested if there still is a relation: 1) First I subsampled each month for each animal. 2) I took the animal with the lowest number of fixes. For each other animal I randomly chose as many fixes as the animal with the lowest number had. I repeated that 20 times. 3) Then I calculated the UD for each of subsampled data. 4) I compared the UD results of those random chosen fixes to the one of the original data which used many fixes. I used linear regression to compare the results. 5) Still there a relation So I do have a problem because there is a relation of the number of fixes and the UD size. Are there any other ideas than subsampling by hourly data? Help would be very much appreciated! Dagmar Am 27.08.2017 um 15:07 schrieb Mathieu Basille: Hey Dagmar, What do you mean by "testing if the UD size is related to the number of fixes"? Did you model UD size as a function of number of fixes using all fixes for all animals? If that's what you did, you could also check if it's true by animal too. One way to do it would be to compute UDs for each animal over samples of fixes for instance every 10, 20, 30, 40, 50 and 60 minutes, and see if there is any relation here. If there is, you may not have other option than subsampling at the hour scale to compare all animals (but what does the UD mean if there is such variation related to sampling? I'm not familiar enough with BRBs to comment here). If there is no relationship per animal, then you simply have animals with high number of fixes AND large UDs! Hope this helps, Mathieu. On 08/23/2017 03:59 PM, Dagmar wrote: Dear all, I want to compare the homerange size between animals based on GPS data but it seems not to work. Here is what I did: I calculated the homeranges using the Biased random bridges from adehabitat HR. Animal_ltraj <- as.ltraj(animalxy, animaltime, id=TierID) D_Animal <- BRB.D(Animal_ltraj, Tmax=21600, Lmin=36) Animal_Grid <- ascgen(allanimals_xy, cellsize=32) # I chose 32 because it was the smallest cellsize chosen automatically. I thought to compare homerange size between animals I must use the same grid for all of them. Animal_BRB <- BRB(Animal_ltraj, D= D_Animal,type=c("UD"),Tmax=21600,Lmin=36, hmin=100,grid=Animalgrid) kernel.area(Animal_BRB, unout=c("km2")) My problem: I have very different sample sizes (i.e. number of fixes): Most animals were located hourly but two animals were located about each 5 min. This results in sample sizes of about 150 fixes per animals and some animals do have almost 10.000 fixes for the same periods of time. Because I want to compare the resulting home range sizes I tested if the UD size is related to the number of fixes and unfortunatly (!!) they are. That means that I cannot compare homerange sizes between my animals My question: - did I do something wrong? - is that problem known and is there a way to solve it? The only way that I find is to reduce the number of fixes of my frequently located animals artificially / randomly to start with the equal number of fixes for all animals. This would be a shame though! Help would be very, very, very much appreciated!!! _______________________________________________ AniMov mailing list [email protected]<mailto:[email protected]> https://lists.faunalia.it/cgi-bin/mailman/listinfo/animov Am 27.08.2017 um 15:07 schrieb Mathieu Basille: Hey Dagmar, What do you mean by "testing if the UD size is related to the number of fixes"? Did you model UD size as a function of number of fixes using all fixes for all animals? If that's what you did, you could also check if it's true by animal too. One way to do it would be to compute UDs for each animal over samples of fixes for instance every 10, 20, 30, 40, 50 and 60 minutes, and see if there is any relation here. If there is, you may not have other option than subsampling at the hour scale to compare all animals (but what does the UD mean if there is such variation related to sampling? I'm not familiar enough with BRBs to comment here). If there is no relationship per animal, then you simply have animals with high number of fixes AND large UDs! Hope this helps, Mathieu. On 08/23/2017 03:59 PM, Dagmar wrote: Dear all, I want to compare the homerange size between animals based on GPS data but it seems not to work. Here is what I did: I calculated the homeranges using the Biased random bridges from adehabitat HR. Animal_ltraj <- as.ltraj(animalxy, animaltime, id=TierID) D_Animal <- BRB.D(Animal_ltraj, Tmax=21600, Lmin=36) Animal_Grid <- ascgen(allanimals_xy, cellsize=32) # I chose 32 because it was the smallest cellsize chosen automatically. I thought to compare homerange size between animals I must use the same grid for all of them. Animal_BRB <- BRB(Animal_ltraj, D= D_Animal,type=c("UD"),Tmax=21600,Lmin=36, hmin=100,grid=Animalgrid) kernel.area(Animal_BRB, unout=c("km2")) My problem: I have very different sample sizes (i.e. number of fixes): Most animals were located hourly but two animals were located about each 5 min. This results in sample sizes of about 150 fixes per animals and some animals do have almost 10.000 fixes for the same periods of time. Because I want to compare the resulting home range sizes I tested if the UD size is related to the number of fixes and unfortunatly (!!) they are. That means that I cannot compare homerange sizes between my animals My question: - did I do something wrong? - is that problem known and is there a way to solve it? The only way that I find is to reduce the number of fixes of my frequently located animals artificially / randomly to start with the equal number of fixes for all animals. This would be a shame though! Help would be very, very, very much appreciated!!! _______________________________________________ AniMov mailing list [email protected]<mailto:[email protected]> https://lists.faunalia.it/cgi-bin/mailman/listinfo/animov _______________________________________________ AniMov mailing list [email protected] https://lists.faunalia.it/cgi-bin/mailman/listinfo/animov
