AIC for analyses of a huge dataset

[Marcos Texeira <mtexeira@ifeva.edu.ar>]

Dear all

I have a question about the applicability of AIC to compare models. My
dependent variable is the frequency of transition among two land covers,
obtained from satellite imagery. I´m trying to correlate this information
with environmental information, using generalized linear models (logistic
regressions).

I have a huge number of trials (2x10^8 trials), and thus even every weak
relationship results significant in statistical terms (but not in
ecological terms). Thus, AIC values are in the order of x10^10. How can I
deal with model selection since DELTA AIC and AIC weights are
meaningless?

Thank you!

MArcos

Marcos Texeira
Cátedra de Métodos Cuantitativos Aplicados / IFEVA - CONICET
4524-8070  int. 8114
Facultad de Agronomía
Universidad de Buenos Aires