www.phidot.org

Hello all,

I am running CJS mark-resight models for an overwinter survival study I am doing on White-throated Sparrows. I am looking at any potential effects of Bird Age (juv vs adult), Sex, or Morph (tan-stripe vs white-stripe) on Phi/p within the species, and any interactive effects of the covariates on survival and resight probability. Total, I have three data sets from two different study sites and seasons. Because each data set has multiple equally competitive models (anywhere from 2 to 9), I was going to use model averaging to get my real estimates for Phi/p. This is where I run into two issues.

1) The model.averaging output from R Mark has 440 or 880 entries depending on whether I run Phi/p averaging separately or together.

2) The output table does not seem to give a clear indication of what real estimate is associated with which covariate or group. The most it gives me is something like "Phi g000 c1 a3 t4" if I run Phi and p separate. If I run them together, it doesn't give me any sort of label for the estimates.

So my first question, is there any way to cut down on the number of entries it gives me to make it more readable? I've tried cutting down my model.collect table to just include the competitive models, but that didn't make a difference.

Second question, is there a way to get the output labeled in a way that's easier to interpret what the effects of the covariates are? I am not having any issue with that if I look at the models individually. For example, one of my top models for one data set has an interactive effect of Bird Age and Morph on Phi. Looking at the outputs from the model, I can see that juvenile tan-stripes have the highest survival estimate, followed by adult white-stripes, adult tan-stripes, and juvenile white-stripes. But other competitive models include effects from Morph alone, Sex alone, and interactive effects of Sex and Morph. Once I do model averaging, it becomes unclear what the effects are and what the differences are between the covariates/groups, as well as what the real estimates are.

Am I just misunderstanding how model averaging is used or works within R Mark? Please let me know if you have any clarifying questions, and thank you for your help and time.

-Nate

There are 440 values for Phi and for p because that would be the number of unique parameters for each with the all different PIMs that RMark uses to maintain flexibility. You don't seem to have a grip on RMark fundamentals so I suggest you need to read more of the documentation. If you ran model.average and specified one of the parameters like model.average(results,parameter="Phi") it will show the design data associated with each parameter. Not sure why you are only seeing the label at the beginning which represents cohort, age and time. You won't see any data if you don't specify a parameter but it gives the par.index which lets you link to design data. Only reason to do p and phi together is if you need covariances between them which you probably don't need. Did you run example(model.average.marklist)? Have you read the workshop notes or Appendix C of Cooch and White? Section 9 of the workshop notes discusses model.averaging. help("model.average.marklist") will show help and the example. It appears that you don't have individual numeric covariates and that all of your covariates are factor variables specified in groups so the values should all be in the design data and you should see them there.