Multistate GOF test limited to UCARE with a c-hat < 1

[Miguel]

I was hoping to get the advice from those of you that are more experienced than me on a somewhat ambiguous situation.

I’m conducting multistate analyses for two species with 120 months of data. Models without transients have very poor fit and thus I based my c-hat estimates on the assumption of a transient starting model (WBWA+3G.Sm+M.ITEC+M.LTEC). 120 periods is too much for MARK to conduct median or bootstrap GOF tests. The c-hat estimates from UCARE are less than one (0.37 and 0.35) I assume because of the large proportion of transients making the data sparse. The output shows “NaN” for most periods which I assume is a result of insufficient data.

Given the lack of a clear course of action when c-hat < 1, I could go ahead and use a c-hat value of one. However, the M.ITEC tests were not significant for males of both species indicating trap-happy males.

I converted the data to single strata and again used UCARE to estimate c-hat for both species. The reduction in the number of parameters allow for more occasions to have sufficient data for the various tests and producing c-hat estimates of 1.43 and 1.87 for both species.

Which c-hat value do you think is most reasonable to use, 1 from the MS or the 1.43 and 1.87 from the single strata estimates?

Your advice is greatly appreciated
Cheers,
Miguel