GOF and Multi Strata Live Dead Models

[kmk]

I am attempting to run a goodness of fit test for a multi strata model. My general model is not a time-dependent model. As such, I assume that based on the instructions in the multi-strata chapter that I should be running a median c-hat test to estimate the goodness of fit. However, when I click on the option for this test in MARK, I am given the reply "The median estimate of c-hat only works for live recaptures and dead recoveries at this time". If anyone has any suggestions regarding this issue or goodness of fit tests for multi strata models in general, they would be greatly appreciated.

[cooch]

I am attempting to run a goodness of fit test for a multi strata model. My general model is not a time-dependent model. As such, I assume that based on the instructions in the multi-strata chapter that I should be running a median c-hat test to estimate the goodness of fit.

That chapter is focused on MS live-encounter data, and thus any recommendations there don't necessarily apply to permutation of the MS framework (e.g., MS live-dead).

However, when I click on the option for this test in MARK, I am given the reply "The median estimate of c-hat only works for live recaptures and dead recoveries at this time".

Which means just what it says -- although you can run a median c-hat on a live encounter-dead recovery analysis, if there is only a single stratum.

If anyone has any suggestions regarding this issue or goodness of fit tests for multi strata models in general, they would be greatly appreciated.

You mean, MS live-dead models. Since there is no good agreed-upon test available in MARK, your first recourse (as suggested in Chapter 5) is to do a manual 'sensitivity' test by iterating chat from 1 to (say) 2 in small increments, and seeing how sensitive your model weights are to c-hat.

[kmk]

Thanks...that explains a lot.

I did have one more question regarding this topic. I noticed that there is an estimated c-hat included in the text-file model output for all the models. Should I assume that this is an incorrect estimate as it is not possible to get c-hat estimates for these types of models? If so, does anyone have an idea what this number indicates? Thanks again for all suggestions/comments.

[cooch]

Thanks...that explains a lot.

I did have one more question regarding this topic. I noticed that there is an estimated c-hat included in the text-file model output for all the models. Should I assume that this is an incorrect estimate as it is not possible to get c-hat estimates for these types of models? If so, does anyone have an idea what this number indicates? Thanks again for all suggestions/comments.

The use of that particular 'estimate' of c-hat is discussed (particularly in the context of the bootstrap) in Chapter 5. See bottom of p. 6 (to start), then read section 5.6 to see how you could use the reported c-hat in the context of a bootstrap approach. I suspect this will help answer your question.