transience

[annaren]

In UCARE the GOF of my data reveals that Test 3.SR is significant for only one of my six groups.
Should I incorporate transience for all groups in the model, only the one or none at all?
If I include make only group 1 contain transients it lowers the AIC by 2.12 compared to if all groups contain transients.
Thanks

[Eric Janney]

I posted a very similar question last Oct. and no one responded. The UCARE manual advocates using a "Time since marking" model on survival if you detect transience via test 3.SR. What I found with my data when I looked at the contingency tables for each period was that test 3.SR tests were significant only in a couple of periods and it varied by group (i.e. sex). So, I asked on the forum whether it was a reasonable modeling strategy to incorporate the TSM effect only on periods or groups with significant test 3.SR results? I discussed the issue with David Anderson a while back. David’s thoughts were that if there is no consistent pattern in the GOF tests and you don’t have a good biological explanation for why there is a lack of fit in some periods or groups, but not others, it is probably best to chalk it up to heterogeneity. At that point, the best approach would be to increase the variance inflation factor (c-hat) to account for the heterogeneity. There are a number of ways to estimate c-hat discussed in the manual.

[annaren]

Thanks for the reply. One of my groups (males) shows significant transience whereas the other (females) does not in test3.SR. However, if I incoroprate transience only in the males it doesnt affect the model that much (AIC only reduced by 0.6). Because I DO have a good biological reason for thinking that there will only be transience in males (as they are the dispersing sex in my species) I included this model in the final set that I then averaged. I hope this is correct.
Thanks again,
Anna