Appropriate to use BIC?

[mtreglia]

Dear All,

I am analyzing mark-resight data from surveys for lizards through nearly all of their habitat on a small island trying to derive a population estimate. There were only 4 survey occasions and I have encounter histories without individually identifiable marks. So I am using a Huggins CC model to estimate population size. The lowest AIC model gives an estimate with what definitely seems like a low estimate (117) with huge confidence intervals (115-5750). The low end of the CI is the actual number individuals marked. This model is p(t)c(t). I wouldn't expect p to necessarily change in time, but I would expect c to, though I know either way is possible. I am not looking at any additional covariates.

From reading the MARK manual and going through Link and Barker (2006) it looks like the BIC may be more appropriate in my case because of the simplicity of the models I am comparing (the only factor being included is "t"). The highest BIC model is p(.)c(t), gives a population estimate of 450, and confidence intervals of about 160-2550. This model has the second lowest AIC, but is obviously drastically different, though seems more reasonable to me.

I know from empirical work that this species has very low detectability, so with 4 surveys I will not get many recaptures/resights and thus get wide confidence intervals regardless.

I did a search through the forum about this topic but did not see anyone using BIC, but it seems appropriate from my understanding, but being relatively new to this I was hoping to get some confirmation (or denial...)

Thank you all in advance.
Cheers,
Mike Treglia

[gwhite]

Instead of worrying about BIC vs. AIC, you better be a lot more worried about your knowledge of the closed capture models. The model p(t) c(t) is not identifiable, and always gives back Nhat = M(t+1). I suggest you do some reading in the gentle introduction, plus the following paper:

White, G. C. 2008. Closed population estimation models and their extensions in program MARK. Environmental and Ecological Statistics 15:89–99.

You just missed the latest MARK workshop, but you would benefit greatly from attending one.

Gary