I am trying to fit some presence-absence data for prairie species in the
incidence function metapopulation model based on the paper by
Quintana-Ascencio and Menges(1996). I am using SPSS and I have also tried
using SAS. However I have been running into the following problems:

1. Some times the data fail to converge even after 500 iterations - this
tells me that the species does not fit the model. Any comments?

2. Sometimes the data converge, but the standard error is greater than the
values of the unknown variables. This is a big problem because that
questions the very validity of the estimate.

3. Sometimes the data converge at two very different points depending on the
parameter values. In this case, the question I have is, which one is
correct?   Is the one with less iteration the better answer, or the one with
a smaller standard error?

If anyone has done a non-linear regression(Levenberg-Marquardt) before and
has some thoughts/comments/suggestions please let me know. Thank you.

Suneeti Jog
Post-Doctoral Associate
Kansas Biological Survey
University of Kansas
Email:[EMAIL PROTECTED]

Reply via email to