Median c-hat and robust design?

[mks]

Hi,

I'm running some models using robust design data types (robust closed captures and also robust design Huggins full het). I'm working in RMARK but I was hoping to use the median c-hat GOF approach in MARK. After importing my models into MARK, retrieving my global model and selecting median c-hat under the tests menu, I get the following error:

"The median estimate of c-hat only works for live recaptures and dead recoveries at this time."

I've read Ch 5 of the MARK book and I was under the impression that the median chat approach could be used with the robust design data type as long as individual covariates were not used. Perhaps I'm missing something obvious but any feedback on this issue would be greatly appreciated.

Thanks,

MKS

[cooch]

Hi,

..."The median estimate of c-hat only works for live recaptures and dead recoveries at this time."

Because it isn't available for the robust design.

I've read Ch 5 of the MARK book and I was under the impression that the median chat approach could be used with the robust design data type as long as individual covariates were not used. Perhaps I'm missing something obvious...

...or not reading carefully.

1\ robust design is mentioned *only once* in Chapter 5, and that is only terms of discussing the saturated model (p. 3)

2\ robust design is not discussed *at all* in terms of GOF anywhere in Chapter 5 (so not sure where your 'impression' is coming from).

3\ what you can test is the fit of your data to an open population model, by (i) collapsing your secondary samples into a single occasion, 1 or 0 depending on whether or not the individual is encountered at least once, or not), and (ii) taking the resulting aggregated histories (which now amount to a simple CJS-type model) and analyzing fit of those data, using median-c-hat.

But, as noted above, there is no full, formal test of fit for a robust design - which is why it isn't mentioned in Chapter 5. Kendall has described approaches based on cell-pooling contingency tests (available in SURVIV, and RDSURVIV as I recall), but I don't believe those are currently recommended.