Pooled model-averaged abundance estimate with CI

[jlaufenb]

You're right Evan, Kate Kendall. John Boulanger and Gary White were co-authors so I had hoped they might chime in as well. If they don't, I'll contact Kate directly. I appreciate the help guys.

Jared

[gwhite]

Jared:

You already were given the answer. Run model averaging on each of the 2 estimates, but simultaneously so that you also get the covariance of the 2 estimates. Most of the models in the set considered had parameters in common across the 2 sexes, so that there is a covariance induced in the estimates. Check the box in the model averaging dialog to output the V-C matrix as a DBF file, which will consist of the 2 N estimates and the 4 values of the VC matrix (where the covariance is duplicated). Then, the SE(N-hat1 + N-hat2) = sqrt(sum(VC matrix)). Use the lognormal confidence interval equation that is in the book, or as Jeff pointed out.

Gary

[murray.efford]

I found this exchange useful - thanks to all. A follow-up detail for Evan: the current MARK book attributes the formula for variance of a model-averaged parameter to Buckland et al 1997, both in 4.5.1 and 14.9.1, but I think the particular version used is that of Burnham & Anderson 2004 (and perhaps 2002 - don't have that at hand to check). Variances from the two equations are different, but usually only by a few percent at most. The later refs are useful as they also give the covariances.

Murray

[cooch]

I found this exchange useful - thanks to all. A follow-up detail for Evan: the current MARK book attributes the formula for variance of a model-averaged parameter to Buckland et al 1997, both in 4.5.1 and 14.9.1, but I think the particular version used is that of Burnham & Anderson 2004 (and perhaps 2002 - don't have that at hand to check). Variances from the two equations are different, but usually only by a few percent at most. The later refs are useful as they also give the covariances.

Murray

Indeed you are correct. When I 'updated' the formulae in various places, I neglected to update the citations. Thanks for spotting it.