To model average or not?

[AndyT]

Hello,

In a normal, recaptures only analysis, my top model (weight = 0.668) has a group effect and the second model (weight = 0.330) is exactly the same model as the top model, but without group. When I do a LR test between the two models, it is marginally significant at an alpha of 0.05 (p = 0.04).

I want reliable estimates of survival, so what I'm wondering is should I average the models, or not? If I average them, the relatively high weight of the second model (without group) means that the difference between the averaged survival estimates will be reduced, since both groups would have the same survival rate in this model. On the other hand, the top model has only twice the weight as the second one, so would it be right to ignore the second one?

Can anyone provide some advice?

Thanks very much!

Andy