Goodness-of-fit test for averaged models

[Kurt van Wyk]

Hi all,

I am looking for a workaround to the following situation.

I am assessing the performance of multiple N-mixture model variants in estimating reptile abundance from spatially replicated DND and count data. My model averaging criterion is ΔAICc < 2. See the example script below.

-----------------------------------------------------------------------

P_glob <- pcount(~MAD+TEMP+RH+TDR
                 ~canopy_cover+scaled_tree_count,
                 data=umf3, mixture="P")

dredged_glob <- dredge(P_glob, rank="AICc")

P_best <- model.avg(dredged_glob, subset = delta <= 2, fit=TRUE)

-----------------------------------------------------------------------

This averaged "best" model is not compatible with Nmix.gof.test(). Is anyone aware of either an alternative way of constructing the averaged model such that it can be tested, or of another appropriate GOF test?

Thank you in advance,

Kurt