fixing parameters on logit or natural scale?

[rasrage]

Hi there,
I need to fix a detection probability to 0 in an analysis. I am not sure whether to actually enter 0 in the "Fix parameter" tab, or whether I am supposed to enter the fixed parameter value on the logit scale (which would be -Inf, but I could approximate it with -99).
When I go through the "Set up numerical estimation run" window under "Fix parameters" and enter either -99 or 0, the results table shows the respective beta parameter as 0. That seems wrong though, because the inverse logit of 0 is 0.5, when I want that probability to be 0.
When I go through the design matrix to fix the parameter, PRESENCE crashes, or rather it doesn't produce the full output.
Any help would be much appreciated.
Thanks!
Rahel

[jhines]

Hi Rahel,

The 'fix parameters' button is for fixing the 'real' variables (eg., psi, p, eps, gam,...), not the 'beta' (untransformed) paramters (eg., a1,a2,b1,b2,...). So, you would fix the p's to zero in the fix parameters window, and make sure the row in the design matrix which corresponds to that p contains all zeros. Fixed parameters are not logit-transformed.

Jim

[rasrage]

Thank you, Jim!