Model averaging and CI estimating in a time-dependent model

[Billy.Requena]

Hi Mark-users,

I collected data regarding a species with exclusive paternal care for the offspring and I am interested in the effect of parental status of males on their survival. I have run my analyses and finished with two equally plausible models (delta QAICc of the second model = 1.1; third model = 9.2). Both models considered survival rate influenced by mean temperature in each month (co-variate used as time dependence proxy), but in one model, males in one parental status consistently show higher survival rate estimates than the other parental status, while in the second model, those estimates fluctuate (during Summer months Status A is higher than Status B, and this situations is the opposite during Winter months). Given I am specifically interested in such effect on survival, I thought in using model averaging procedures. Although I have read about the calculations of averaging estimates and variance (in Burnham and Andersson 2002 book, for example), I am a little bit confused with this time dependence term in my analyses.
Should I just calculate the averaging estimated beta parameters and multiply them by the temperature in each occasion? Or should I perform the averaging procedure using the estimated value of survival in each occasion?
The reason I am wondering about this is that the uncertainty around the estimates changes depending on the sampled month (and consequently, on the co-variate temperature), and I don't think I could capture this using an averaging directly on the beta parameters, right?
Thanks for any help

Gustavo Requena
PhD Student, Universidade de Sao Paulo - Brazil