Fluctuation in Beta values across models

[Justine33]

Dear all,

I am relatively new in using Presence and this may be a simple question. I looked over the previous posts and couldnt find an answer that could clarify my concern.

I am running simple single season on my species using PRESENCE 7.8. My data is quite scarce and I have a detection probability of 0.13. When modelling occupancy across different combinations of 3 covariates- my top model's beta values fluctuate much higher then my other models. Why I'm concerned with this as this obviously influences my averaged beta estimates across all models.
So my question is:
1. Should I be concerned that my best model's beta values are much higher then all my other model beta values?
2. Is averaging beta values across all models good practice when one model has beta values that varies largely from the other models? Even if its the best model.

AIC ranked models:
Model 1 beta values:
Beta intercept: 33.75 (10.52)
Covariate 1: -12.71 (4.31)
Covariate 2: 20.76 (6.27)
Model 2 beta vlaues:
Beta intercept: 2.07 (1.19)
Covariate 2: 1.86 (0.91)
Model 3 beta values:
Beta intercept: 2.1 (1.15)
Covariate 1: 1.72 (0.79)

Any help clarifying how to interpret this would be helpful.
cheers.

[jhines]

With sparse data and low detection rates, occupancy estimates will frequently go to 1.0 and all of the variation in the data will be taken up by the detection estimate. I don't know what your covariate values are, but my guess is that the 'real' occupancy estimates from the top model will have many sites with occupancy estimated to be 1.0. So, I think your top model is probably too complicated for your data and you should stick to only one covariate at a time. How many sites and surveys do you have. I believe there is a thread on this site somewhere which gives a 'rule of thumb' about how many covariates can be used for a given number of sites and surveys.

Jim

[Justine33]

Thank you Jim for the response.
Yes indeed I only have a few sampling sites(60). Ill try and reduce my covariates.

cheers.