Order of parameters in Robust models and testing for closure

[Miguel]

In the Gentle Introduction's examples of closed models, the p’s and c’s are ordered separately so that design matrices are easier to construct and beta parameters are interpretable. However, in Robust models the order of parameters alternates between p’s and c’s and in mixture models the pa, pb and ca, cb parameters are also not ordered separately. This would make for very odd matrices and uninterpretable betas. Did I miss something? I have more than 165 periods, is there some way to reorder parameters easily?

I have also been looking for a way to test for closure during secondary periods without having to conduct >200 separate closure tests. As far as I can tell, both Capture and CloseTest do not handle closure tests on robust data. Any suggestions would be greatly appreciated.

Thank you,
Miguel

[cooch]

In the Gentle Introduction's examples of closed models, the p’s and c’s are ordered separately so that design matrices are easier to construct and beta parameters are interpretable. However, in Robust models the order of parameters alternates between p’s and c’s and in mixture models the pa, pb and ca, cb parameters are also not ordered separately. This would make for very odd matrices and uninterpretable betas. Did I miss something? I have more than 165 periods, is there some way to reorder parameters easily?

Not sure if you 'missed something', since I don't have a clue what you're talking about. For robust design, most of the time the interest is on estimation of 'open period' parameters (gammas, survival), so more often than not, the model for the closed samples is simply set to M(t) or M(tb), and left that way for all of the closed sample periods.

[Miguel]

Evan, thanks for replying to my question.

Yes, the parameters of interest are the gammas and survival but what if a CJS analysis shows that capture rates are influenced by environmental conditions? Would it not make sense to include a climate covariate for P and c since it might improve model fit and the estimates of the gamma and survival parameters?

Any suggestions about testing for closure?

Thanks again,
Miguel

[cooch]

Evan, thanks for replying to my question.

Yes, the parameters of interest are the gammas and survival but what if a CJS analysis shows that capture rates are influenced by environmental conditions? Would it not make sense to include a climate covariate for P and c since it might improve model fit and the estimates of the gamma and survival parameters?

CJS analysis shows environmental covariates influence things during the 'open' periods, but that doesn't say much about p and c, which are parameters used to model encounters during the closed periods. I'd be inclined to use environmental covariates to model gammas and S, and open p, but not bother doing much that is 'fancy' within closed periods.

Any suggestions about testing for closure?

Thanks again,
Miguel

See Bill Kendall's comments in Chapter 15 on 'closure'.

[Miguel]

Thanks for the help, your points are well taken.

Cheers,
Miguel