Kendall emigration with RBD

[DanG]

I am analyzing MRC data that was collected from a sampling design that called for sampling only a random selection of strata with in the larger study area during each capture event. I have fit this data to Pollock’s robust design and found that data is too sparse to reliably estimate the migration parameters.

Making the assumption that the animals do not move in and out of the study area to a significant degree, I hypothesize that a valid analysis would be to fix the emigration parameters to the known probability that a stratum is not randomly selected and the immigration parameter to the known probability that a stratum is randomly selected. So, again, assuming there is no significant animal movement, capture probability would be, in theory, a non-bias estimate of conditional capture probability (conditional on having the strata selected where the subject resides).

I have never seen this approach taken before and would like any comments on the it’s validity and/or recommendations.

Thanks,
Dan

[Bill Kendall]

Dan,

This is an intriguing idea. I would probably need more information in order to be helpful, but let me see to what extent I understand what you presented.

Of the target population you are interested in, you can only set up capture effort in a subset of the area at each sampling occasion? You choose areas completely randomly, with replacement across time (e.g., you choose 5 out of 10 sub-areas each time, and sub-area 3 could be chosen in consecutive sampling periods)? Next, is this truly a multistate situation, or are you viewing it that way just because you are imposing an unobservable state? If the latter, do you believe your conditional capture probability should vary spatially (i.e., you would like to consider each sub-area separately)? Most importantly, what demographic parameters are you most interested in?

Because my answer would vary depending on the answers to these question, I believe I'll leave it at that.

Cheers,
Bill