www.phidot.org

Hi,
I'm trying to do quadratic (or higher power) NestAge nest survival models in RMark. In regular MARK, the process is described in the MARK Book (Chpt. 17, page 15, last paragraph), but I couldn't find it in the RMark appendix. Here is what I tried, but I'm definitely not convinced that it worked.

Code: Select all

gwage2=mark(gw,nocc=100,model="Nest",model.parameters=list(S=list(formula=~NestAge+I(NestAge^2))))

Here is the result it produced:

Code: Select all

              estimate        se        lcl       ucl
S:(Intercept)  1.5634772000 0.0000000  1.5634772 1.5634772
S:NestAge      0.0002934439 0.0450713 -0.0880463 0.0886332
S:I(NestAge^2) 1.5634743000 0.0000000  1.5634743 1.5634743

Any help would be much appreciated!

Mike
New Jersey

Your intuition is correct. If you look at the design matrix that it created you'll see that the first row looks something like

1 NestAge1 1

That is why the first and third betas have the same values.

NestAge is a weird one that I added for this particular model. It is actually computed as a time-varying individual covariate when it builds the model. I didn't create any capacity to use functions of it so unfortunately the answer at this point is that there is no way to do it in RMark. Given that the nest models can be fit with standard statistical software I'm not likely to expend much effort programming on it.

regards --jeff

Thanks Jeff. Not a problem to use regular MARK. When you say standard statistical software is that what you're referring to? I was under the impression that MARK's treatment of nest age and date of season variables for nest survival is a little superior to other methods in SAS or regular R (i.e., logistic exposure method).

Mike

Mike-

No I was referring to SAS,R or whatever. I was unaware that MARK was superior to any of those. I know there is some published literature on this subject. Maybe others can chime in that may know more than I do about the subject.

--jeff

The likelihoods from these other packages are generally identical to what you get in MARK unless you have time-varying individual covariates, and then there can be issues if not handled properly. I think the main advantage of MARK is the ability to easily model average estimates, but if you know how to program these other packages....

Gary