Hi all,
I have data for only 3 primary periods that is being modelled in the robust design, and I want to be able to do a goodness of fit test and estimate the c hat. I have pooled my data in secondary occasions so that I can run a CJS model to do GOF in U-CARE, but as I only have 3 primary occasions, it can only give me information about transience and not trap dependency. Transience is significant, as expected.
I then thought that I could use this same CJS in MARK and run the median c hat on the data. So on the full time dependent model, I get under dispersion in my data. But if I then try to use the median c hat on the model that includes survival modelled with time since marking to see what the improved fit would be. It comes up with an error saying
'Every one of your simulated values generated a c hat less than the observed c hat value. Logistic regression cannot be performed'.
The reason I am questioning this is because I have also run the data of the Robust Design in Program RDSURV and based on the most general model I get a c hat that is larger than 3. So I am a little confused in which of the methods is best for me to be using and how I should proceed with modelling in the Robust Design and I want to correct for any over-dispersion.
Any thoughts would be greatly appreciated!
Thank you