SE = 0 for RDMultScalOcc

[rcscott]

Hi all!

I am running in to a small problem. I ran a dynamic multi-scale occupancy model on a relatively small set of data on the amphibian chytrid fungus. I would like to plot occupancy of the disease in each site by year. However, b/c the data are quite sparse, the beta variances are negative and the standard error defaults to 0. Is there another formulation for calculating SE that could work? My supervisor mentioned using a profile likelihood, but I'm not sure how to implement that in RMark?

Thanks!
Reed

[gwhite]

This sounds to me like you do not have sufficient data to estimate some of the beta parameters. If you want me to verify this assumption, send me the output to look at.

Gary

[rcscott]

I talked with both Gary and Jeff about this off list, both were incredibly helpful. The answer is that because the data were very sparse, only 3 out of 5 parameters could be estimated. Because of this the variance-covariance matrix is singular, so the standard error is 0.