from CJS to Pradel?

[Claudio]

I'm analyzing CMR data on dragonflies. I have 2 grouping variables, sex and
presence of parasites. I would like apply Pradel model to asses survival and
recruitment rates. I have dropped out several models without biological
meanings on the basis of a priori hypotheses. Nonethless, I still have a lot
of models to run.
My question is: does it make sense start with CJS models, then fit in the
Pradel framework only the best CJS models (of course, extended with
recruitments)?
Thank you in advance
Claudio Angelini
Sezze

[abreton]

If you are interested in estimating recruitment then it does not make sense to even consider the parameterization (phi, p) specified by the CJS model. And note that even if you did fit your data to the CJS and Pradel reverse-time/recruitment models, the AIC values would not be comparable since the likelihoods specified by these models are inequivalent. Similarly, AIC values from the Jolly-Seber (includes recruitment) and Pradel models cannot be compared - see the MARK help file under Pradel > Recruitment Parameters for more information.

I wonder if you wrote 'CJS' but meant 'JS' (Jolly-Seber)? Certainly, the Pradel model discussed here and JS are in the same class of models so you should consider the strengths and weaknesses of each of these given your data along with others including the Link-Barker parameterization (see losses on recapture under Pradel > Recruitment Parameters in MARK help file). I recommend that you read through Chapter 18 in Williams et al. "Analysis and Management of Animal Populations" and Chapters 13-14 in a Gentle Intro to Mark if you haven't done so already. Rather than use the best JS model as a template to build our starting Pradel model (e.g., sex but no presence fitted to survival), it seems more logical, given the important differences between these two models, to decide which model type (from this class of models) is best given your data and then use it exclusively in your analysis.

[Eric Janney]

I think Claudio is asking is if it is OK to use the same structure on phi and p from his top CJS model as a starting point for modeling phi, p, and f in a Pradel model. Then he can simply focus on modeling f. If that is indeed what your asking, I think that it is a good starting point. However, you should realize that the pradel model has to estimate additional parameters for recruitment. So, you may want to try less structure on phi and p than what was selected as the top CJS model and see how it ranks with AIC.

[Claudio]

Thanks for both your answers. Yes, Eric, you are right. I did not want to compare CJS versus Pradel models, but use the CJS models as a kind of preliminary test. Thus, it does make sense. Thank you again.