Time varying habitat covariates

[geoffwah]

Hi All

I'm running a multi-season occupancy analysis in RMark, in which the focus is estimation of the effect of several habitat variables on epsilon and gamma. The problem I have is that I would like to incorporate annual changes in some of these variables (over the six year study, I have estimates of some variables in each year). Hence, I want the relationship between epsilon/gamma and these variables to incorporate these changes.

In a post a couple of years ago, Jim Hines informed that this was possible in Presence, by setting the time-varying habitat covariates as sampling covariates. You can view the thread here: viewtopic.php?f=11&t=1134&p=3054&hilit=jim+multi#p3054

Is it possible to do this in RMark? If so, how might I go about it?

I would greatly appreciate any advice!

Regards

Geoff

[darryl]

Hi Geoff,
I'm a bit unclear on the model you're actually wanting to fit to the data. Is it gamma(Habitat), gamma(Habitat+time) or gamma(Habitat*time)? Then, does the habitat value itself change over time at each sampling unit? What I'm not sure about is whether you want the relationship to change each year (ie some year a high habitat value is good, other years it's bad), or the relationship is constant, but the values at a unit change through time (ie higher values are always better, but in some years a particular unit may have a high value, other years it's low).
Cheers
Darryl

[geoffwah]

Hi Darryl

The model is simply of the gamma(Habitat) variety. So yes, I'm assuming the relationship is constant through time, but want to incorporate changes in the variable at each sampling unit through time. Hence, for a particular site, vegetation cover might be 20% in year 1, 25% in year 2, 50% in year 3, and so on.

Cheers

Geoff

[jlaake]

See ?RDOccupancy for an example. In your case, in your data file the individual covariates should be named Habitat1,Habitat2,...,Habitatn where the seasons are numbered 1 to n. Then when you use gamma=list(formula~Habitat) it will know to use individual covariate Habitati for season i. If the time-varying covariate varies by secondary occasion as well then the covariates need to be named covij for the jth occasion of the ith season as in the RDOccupancy example.

--jeff

[geoffwah]

Hi Darryl and Jeff

Many thanks for your help. Jeff, I fit the models no probs withe the approach you described. Lovely!

Cheers

Geoff