model selection of different parameterizations

[Diego.Pavon]

Hello,

I am currently running different parameterizations of the same data set.
Is it a good approach to run the "basic models" (i.e. with no covariates) in all the parameterizations to see what kind of parameterization shall I use for the covariates models? And also, is it good to run within the same project, different parametirations and then compare AIC, weights, and so on? Or does the program get a bit confused when using different parameterizations in the same project? If so, how can I compare between models to do model selection?

Another thing, is it correct to use the first parameterization in PRESENCE with 0's in extinction and colonisation? These models would represent no dynamics, am I right? And then I would assume a constant occupancy before hand?

Thank you very much

Diego

[jhines]

Without covariates, the different parameterizations should give the same results. The reason for choosing one parameterization over another is how you wish to model the parameters as functions of covariates. If you think colonization or extinction depends on a particular covariate, then the default parameterization is fine. However, if you wish to build a model where seasonal occupancy (psi) is a function of a covariate, but colonization and extinction are not, then you would choose the 2nd or 3rd parameterization listed. The 4th parameterization is a slightly constrained model, with colonization set to be equal to 1-extinction. Another factor in choosing a parameterization is model convergence. The default model is most likely to converge, where the others will sometimes have problems, due to sparse data. So the best choice is the default parameterization, unless you wish to model seasonal occupancy as a function of a covariate. There is no reason to run the other parameterizations, unless you wish to check that you get the same output for different parameterizations.

Yes, if you set all colonization and extinction parameters to zero, you'll have a model identical to the single-season model (constant occupancy).

Jim