POPAN pop estimates - time-varying covariates

[SoConfused]

Hello,

I'm running a few POPAN models; the models use age as a time-varying covariate (I have multiple initial ages). When I get to using popan.derived to calculate pop abundance by group/occasion, I get the error "Computations for mlogit parameters with covariate values cannot be specified with index column in data; use separate indices argument".

It looks like I can get the group / occasion N-hat values if I load the RMark results into R. However, I can't figure out how to properly sum N-hat estimates across groups within occasion (since I'm not sure where the vcv matrix is hiding). I also thought of exporting the models into MARK, re-running them as mcmc models, then follow this post: http://www.phidot.org/forum/viewtopic.php?f=1&t=3744 to re-export the results into R and sum them up that way. But that still seems unreasonably complicated.

Is there anywhere a worked example of how to calculate N-hat values with individual covariates?

Thank you!

[jlaake]

Please search the list before you post a question. You should NOT be using individual covariates with a Jolly-Seber type model (eg POPAN). See http://www.phidot.org/forum/viewtopic.php?f=21&t=3631&p=11877&hilit=popan+individual+covariates#p11877.

But are you really using an individual covariate or did you use age as a group variable? If the latter then it should work.

[SoConfused]

Yes, I realized my mistake the other day (should've deleted this post). I was using age as an individual covariate, but now changed over to using age in grouping, and all's well.

One follow-up question though. My recap depends on a grouping variable and on age (as a linear function). How do I correctly use popan.derived with age-structured recap values? The help files state that popan.derived uses covariate.predictions, and I remember reading somewhere (although can't find now) that indices need to be supplied separately. I've used covariate.predictions before, but can't figure out how to work through popan.derived to account for ages changing over time.

[jlaake]

You need to understand that I wrote popan.derived because I needed the abundance estimates for an analysis I was doing, and at that time I was unable to extract the derived estimates in the binary file from MARK. That has been fixed and you can get the estimates of abundance by group that is in the MARK output file in model$results$derived where model is the name of the mark model object.

There are a few extra items that I provide in popan.derived that aren't in the MARK output file but unless you really need them then I suggest that you use what MARK provides.

[SoConfused]

Good to know, thank you. In that case - is it correct to assume that MARK correctly accounts for age-structured recaps when calculating derived values? That is, that ages are correctly propagated between sampling events (as opposed to using a mean value per group, for example).

[jlaake]

I'm not entirely sure why you are confused. But as long as you assign the correct initial age for each group, the correct time intervals and age unit ( default is 1) then RMark assigns the advancing age over time . You can check those values by looking at the design data for each parameter. Then if age is used in formula for p and Phi in the model these are used in estimating N by group and time. Hopefully that answers your question.

[SoConfused]

Perfect, thank you!