N-hat for huggins closed capture models

[bmitchel]

Hello Phidot,

How does MARK calculate N-hat for closed capture models with individual covariates? According to the help, for models without individual covariates, N-hat = M(t + 1)/[1- (1-p)^n], where M(t + 1) = number of individuals known to be in the population and n is the number of occasions and p is the estimated detection probability. The help says that "a more complex estimator is required" for individual covariate models. I thought that I could substitute the logit link of the parameter estimates times the covariate values reported in the output for p, but this didn't match my MARK output.

Also, I noticed the reported confidence interval is not symmetric. Is this a likelihood-based CI, or is there some transformation of N-hat and SE N-hat?

Thanks!

Brian Mitchell

[gwhite]

Brian:
N is computed as the sum over all M(t+1) animals of 1 divided by p*-hat for each individual. That is, each individual has its own p* value, computed for its set of individual covariates. Go see Huggins papers for more details.

Confidence intervals are computed based on a log transformation, so are asymmetric.

Gary