Another variogram inspection question

[Rebecca Leloudis]

Hi everyone,

I'm joining in on having difficulty evaluating range residencies in my variograms, and I also have some concerns about the top models selected by ctmm.guess and ctmm.select for some individuals. Apologies if these questions have been addressed, I think I mostly just need second opinions on if I'm evaluating variograms correctly or not.

In some cases, like Arwen and McMuffin here (please forgive our silly naming systems!), I'm unsure if they've truly reached an asymptote. The best fitting models selected by ctmm (Arwen = OU anistotropic, McMuffin = OUF anisotropic) also overlap very little with the empirical variograms. Between that and the unclear asymptotes, I'm not sure either are range resident enough to use in my analysis (I'm estimating home ranges and evaluating resource selection therein).

Others, like Gorthaur and Lobelia, approach an asymptote but deviate at the end--my understanding is that this is just a natural byproduct of the variance between observations at lag = 0 and lag = x months, but I wanted to confirm.

And finally, some (like Frostbite) simply have confusing patterns where I'm not sure if I should consider them range-resident, or simply wait until more data is available and decide from there.

I should note that we have used varying fix rates in this study, all of which are fairly coarse. Our finest resolution is an hourly fix, but our study species, fisher, lives in tree cavities and therefore we have a lot of gaps in our data. I did include the different intervals when generating variograms.

Thanks so much!

[Christen Fleming]

Hi Rebecca,

Lobelia looks great. The end of the variogram is largely noise and doesn't mean anything. At the other end, Arwen and  McMuffin show moderate discrepancies.  McMuffin has a slow climb in the variogram consistent with a slowly shifting range. Frostbite and McMuffin appear to have returned back after a year's cycle. You have to decide if you want to estimate seasonal ranges or lifetime ranges. If you average together better-fitting seasonal ranges to get the lifetime range, the result won't differ from using all of the data unless the discrepancy becomes large, and I'm not sure if you're at that point for any of these.

Best,

Chris