GOF when deviance=0

[Renata]

Hi MARK-users,

I'm running a CJS analysis on a bird dataset in which I used dots to code some missing data points. That works great, but I always get deviance = 0 for all models. What seems to happen is that the -2logL of the saturated model is always higher than for any other model, so the deviance is actually negative, and it is shown as zero.

Reading postings regarding cases when deviance = 0 for other reasons, I suspect that what it is happening is that the -2logL of the saturated model in cases where missing data is coded with dots (a recent implementation in MARK) is not yet properly calculated, but that the AIC values should still be okay, so I could go ahead and use those results.

That's good news, but how can I conduct a GOF analysis? Since the deviance is 0, the c-hat is also zero, so I can't use those, and RELEASE and U-CARE don't like the dots...

Any tips are very welcome, thank you!

Renata
rrd32@umsl.edu

[hpdutra]

Oi Renata sumida!!!

Have you tried bootstrapping it. Section 5.6 in "Gentke introduction"?

Beto