problem with psi estimates over 1

[chilesummits]

Hello everyone,

I'm currently running single season models with several covariates. I have 9 sites and 3 different study areas with 113 days of sampling each. I've run models with 2 to 5 parameters, the naive estimates vary from 0,5 to 0,9 between the 3 study areas, but all my covariate psi estimates are very large, ranging from 1 to 100 or more in some cases, and with negative p estimates as well.
The overall p is very low (under 0,05) so I grouped survey occasions into 5 day periods instead of 1 day, to increase the p. Now it is close to 0,15 but the covariate psi estimates are still over 1.

I appreciate your time and any help I can get.

[jhines]

You should never get estimates of psi or p which are less than zero or greater than one. (If you do, I ask that you send me the input data.) Also, the estimate of psi should be greater than or equal to the naive estimate.

So, I assume you are referring to the untransformed 'beta' estimates of occupancy or detection. Those values can range from minus infinity to plus infinity. If a 'real' estimate of occupancy, psi, is near 1.0, then the corresponding untransformed 'beta' estimate can be very large. 'Real' parameter estimates near zero can be very large negative numbers, and any 'real' estimate less than 0.5 will have an untransformed estimate which is negative.

Beta estimates associated with covariates can be negative or positive, depending on whether the covariate has a positive effect (higher values of the covariate correspond to higher values of the parameter) or negative effect (higher values of the covariate correspond to lower values of the parameter).

Jim