Error with Huggins Pledger models

[gwhite]

I've discovered a problem with the Huggins Pledger mixture models, where the conditioning on the never-seen category was done independently for each mixture rather than jointly. The new code corrects this problem, but is going to change estimates of pi and the derived population estimates. In most cases, the estimates of the p's are the same under both the old and new parameterizations, with changes only in estimates of pi and the derived N. But, I have also found datasets where the new parameterization does not converge to reasonable values. An example of this inappropriate behavior occurs with either of the Carothers taxicab datasets distributed with the program. In contrast, the Edwards-Eberhardt cottontail data work fine with the new parameterization. I have posted a new version of MARK that computes what I now believe is the correct approach. Let me know of problems you have with this new implementation.

Gary

[cooch]

The new version Gary just posted is also mirrored here:

http://www.phidot.org/software/mark/download/