rsf.fit()

[Vitek Jirinec]

Hi Chris,

I hear that more info should be coming up about rsf.fit(), including perhaps a paper and a vignette. Before that happens, I wanted to make sure I'm on the right track when trying to estimate habitat selection coefficients for a population of animals. I'm using info from your seminars and the available ctmm documentation.

My general process:

habitat.rasters <- list(raster1, raster2, etc) # aligned rasters in projection equal to location data

RSF.1 <- rsf.fit(animal1.data, UD=animal1.AKDE, R =  habitat.rasters , debias = T, level.UD = 0.95)
RSF.2 <- rsf.fit(animal2.data, UD=animal2.AKDE, R =  habitat.rasters , debias = T, level.UD = 0.95)

RSF.list <- list( RSF.1,  RSF.2)

RSF.population <- mean(RSF.list)

Are the coefficients in "RSF.population" equal to the average of the separate outputs of rsf.fit? That is, could those be considered the population habitat selection betas?

Thanks! This is exciting stuff.

Vitek

[Christen Fleming]

Hi Vitek,

Yes, we have one paper in review: https://www.biorxiv.org/content/10.1101/2022.04.21.489059v1 and a couple of more papers coming out shortly that give all of the technical details and spell out the advantages.

For the current default numerical method, the rasters don't have to be in the same projection as the data, and would preferably be unmodified/unprojected.

Yes, mean() uses a hierarchical model to estimate the population habitat selection parameters (and the variance among individuals).

If you run summary() on RSF.population then you will get both the population mean parameter estimates and estimates of population variation in those parameters.

Best,

Chris

[Vitek Jirinec]

Thanks Chris & Jesse,

This seems to be working nicely for getting the population mean parameters for our 35 animals with 5 habitat rasters. In my experience—aside from the statistical advantages you mention—coding the RSFs with your approach is much more streamlined than what I have used previously.

Looking forward to the papers,

Vitek