Question about BC Overlap Significance Testing

[Jeremy]

Hello,

I have been working with some data that investigates site/home range fidelity between years.  I have data from 8 animals for two years (May-October) and I am interested if animals are using the same home range areas between years.  Most of my Bhattacharyya coefficient (BC) overlap values are high but I was wondering if there is any significance tests that might be used in combination with the BC values that would determine if home ranges were statistically similar or different between years.  I might be thinking about this in the wrong way and any suggestions would be much appreciated!  

[Christen Fleming]

Hi Jeremy,

If you mean like is PDF_1 significantly different than PDF_2, I tried to construct a formal test and in real data the answer always seemed to be that if you had enough data then they will be found to be significantly different. I think it is unavoidable for reasons of dimensionality. That kind of significance test combines the effect size with data quality in a way that isn't teased apart. And data quality will vary by individual, so that half of the equation isn't fixed. For those reasons I abandoned that research.

The BC quantifies the magnitude of the difference between the PDFs (though its CIs in ctmm are currently calculated under an assumption that the two home ranges are different and can't be used to reject overlap==1 as a null hypothesis even if you wanted).

I would frame my analysis to focus on quantifying how similar/different the home ranges are rather than whether they or not they are exactly the same. None of them are likely to be exactly 100% the same from year to year. You'd want to report something like "X individuals exhibited >90% year-to-year site fidelity (at 95% confidence)".

Best,

Chris

[Jeremy]

Hey Chris,

Thank you for the quick reply!  I really appreciate it!  I think I will use your suggestion of reporting that X individuals exhibited >X% inter-annual site fidelity.  Most of the animals had overlap values >0.8 from between years which seemed to indicate to me that there was a relatively high degree of site fidelity between years.

I had one more quick question I was hoping to ask.  I am also hoping to use the ctmm package to calculate a population level AKDE home range within our study area.  Our animals are limited in movement by a high-fence and I hope to use the overlap function to test to see if our population is using the same areas of the property from year to year.  I am working with data from about 16 animals in both 2015 and 2016.  About 4 of these animals where collared in both 2015 and 2016, but most were unique animals.  GPS location intervals ranged from 15 minutes to 2 hours between fixes. Would you suggest trying to pool all of the GPS locations into two groups (2015 and 2016) to create population level AKDE estimates for each year, or do you think it would be best to create home range estimates for all 32 animals separately and then average the BC values that compare animals between years (2015 and 2016)?

Thanks again for the assistance!

Jeremy 

[Christen Fleming]

Hi Jeremy,

Currently, if you want to make a population UD, you do the individual-level analysis first on the same grid and then feed those results into the mean() method. However, I don't have that coded for overlap() to work, because you need some movement model parameters to get the debias & CIs, which you don't have for the group. (In the future, this will all be done with correlated/hierarchical models on the groups).

Getting overlap values for every pair, within and between groups, would work. You could then compare the overlap distributions. There would probably be more information in that.

Best,

Chris

[Jeremy]

Hey Chris,

Thanks again for the assistance!  This has been a huge help.

Thanks,

Jeremy