Fewer collected points causing larger ADKE range estimates

[Zoe Bonerbo]

Hi Chris,

I'm wondering if you have any guidance on how to deal with estimating home ranges for individuals with fewer fixes. We have GPS transmitters on 20 falcons that collect fixes at different rates depending on the behavior. Most of our birds have thousands of points while several only have a few hundred. When running the one's with a few hundred fixes through variograms they seem to look okay (they plateau out and look like the ones with thousands of points). Then, when estimating the AKDE their ranges seem to be overestimated and "balloon out". We noticed this after converting it to an sf and visually plotting it on a map, where the territory ranges were very exaggerated compared to individuals with thousands of points and transmitter tracks with similar movement patterns in similar regions. I'm assuming there's some level of uncertainty that may be causing the inflation but is there any good way to account for it?

Cheers,

Zoe

[Jesse Alston]

Hi Zoe,

If the variograms look OK, it's very unlikely that the sizes of the home ranges are being overestimated. I think you likely have a relatively small effective sample size for these individuals with fewer fixes, so just looking at the actual locations, there is a lot of area within the home range estimate that does not have locations. If you were to keep collecting data for those individuals, you will see more and more of this area containing locations.

When effective sample sizes are small, the bigger danger is actually underestimation of home ranges. When this is a concern (typically, an effective sample size < 10), you can use bootstrapping to improve the estimate.

For more info, you can read these papers:

https://doi.org/10.1111/2041-210X.13786

https://doi.org/10.1002/ecm.1344

https://doi.org/10.1101/2022.09.29.509951

Jesse

--
You received this message because you are subscribed to the Google Groups "ctmm R user group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ctmm-user+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ctmm-user/f0e0f934-9b47-461f-bee5-f5b5d2f0e5a8n%40googlegroups.com.

--

Jesse Alston

+1 252 532 5236

jmalston.com

[Christen Fleming]

In addition to what Jesse said, you could sub-sample the better tracks and see if the sub-sampled home-range estimates appear similarly "ballooned out" in comparison to the sub-sample versus the entire dataset. I.e., does there actually appear to be positive bias under cross validation.

Best,

Chris