I'm running occupancy analyses (MacKenzie models), trying to estimate psi as a function of individual covariates.
For ANODEV, how should I specify my global model? Would this just be a model that includes all possible covariates?
I do have a group variable, corresponding to 35 different study areas (35 different 3x3-mile landscapes), so I suppose I could specify a {p(grp*t)psi(grp)} model as my most global model. I think this would make good sense to test my models that include landscape metrics as covariates (where each 3x3-mile landscape has a different value for each landscape metric). But I'm not sure this makes sense for testing models with site- or patch-level covariates because the group-model still only provides mean psi for each landscape.
Perhaps a model that includes a group effect (to address landscape-level variation) AND all possible site- and patch-level covariates would serve as my global model?
Any thoughts?
thanks,
jeff