Multiseason w/covars - matrix inversion error

[Kate]

Hello,
I have 2 years of bird data with 3 survey replicates each year. I am attempting to model occupancy and detection as a function of habitat covariates (using 2nd parameterization for seasonal Psi). The covars are seasonal; they change each year but do not change between each survey rep. Following suggestions in previous posts, I entered each year's covar values as a separate site-specific covar and set up the design matrix as:
a1 a2
Psi1 1 HabYear1
Psi2 1 HabYear2

When I model Psi as a function of any covar (with constant colonization, and seasonal detection) I get a "matrix inversion error" message in the results. (I also get this error when entering the covars as survey-specific with year 1 values in column 1, year 2 values in column 2, and 0s in the remaining columns, as also previously suggested). I don't understand what this error means, or how valid the results are in light of this error. Have I set up the design matrix incorrectly? Any suggestions are much appreciated!
Thanks,
Kate

[darryl]

Hi Kate,
The second and third parameterizations can be unstable, particularly when using covariates. This has been discussed a few times over the years here. For a couple of recent posts and ideas of things to try see:
viewtopic.php?f=11&t=2041
viewtopic.php?f=11&t=2028

That particular error usually comes about because some of your estimates are on the boundary of allowable values. Are any of your beta parameters suspiciously large or small?

Cheers
Darryl