GOF testing for nest survival models

[jkemp]

I notice that neither Dinsmore et al (2002) nor Jay Rotella's Chapter 14 address the issue of GOF testing for nest survival models. Is there a reason for this that i'm missing? A currently favoured method?

[gwhite]

Like the known fate data type, the saturated model for the nest survival data type is a useful model. Because the saturated model fits the data perfectly, there is no GOF test.

To help you understand this point, consider a simple radio-tracking study where 100 radios are put on a single age/sex class for one occasion. The saturated model is the simple survival estimate based on the binomial distribution. There is only one data point, hence one degree of freedom, and that d.f. is used to make the estimate of survival. NOBODY ever talks about a GOF test for this situation -- it is obvious that there is no GOF test available.

To obtain a GOF test, we have to assume a reasonable biological model that is reduced from the saturated model. This selection can be pretty arbitrary.

Gary

[cooch]

Thanks Gary -

Just added that text to Chapter 14 (which has just been uploaded).

[jkemp]

Thanks very much Gary,

Josh