Negative lower confidence intervals for derived estimates

[ehileman]

Hi all:

This is related to a previous post found here: http://www.phidot.org/forum/viewtopic.php?f=1&t=969#p6895.

My Popan model averaged derived N-hat estimates have lower confidence intervals that are negative. I suspect that this is due to sparse data and perhaps a lack of asymptotic normality. Therefore, I am wondering if there is a more appropriate method, given my data, to calculate confidence intervals. Can anyone tell me if the method described in CH14, pg 33 of 'the book' (11th ed) (or another method) is appropriate to estimate CIs for open abundance estimates like those in Popan? Williams, Frederick, and Nichols 2011 applied this method to calculate confidence intervals for the superpopulation. From what I was able to gather, this method assumes a log-normal distribution and forces the lower confidence interval to be greater than Mt+1. In applying this equation to N-hat estimates, I used the number of unique individuals capture per occasion rather than Mt+1. Any suggestions would be greatly appreciated!

Eric

[ehileman]

Thanks to Evan for responding (on another thread) to queries related to this post. In the event others are interested in this topic, that exchange can be found here: http://www.phidot.org/forum/viewtopic.php?f=1&t=969#p6951.

Cheers!

Eric