Adding factor covariate in a multiple-season

[Diego.Pavon]

Hi all!

I was wondering if occupancy models may allow me to compare three periods. I mean, I want to divide the study period into 3 sub-periods (beginning, mid, and end of the study period) and compare the occupancy and detection probabilities between them. I would like to campare whether the co-occurrence between two species is different between these 3 periods. Is that possible? Can I add a covariate, called period, which is a factor with 3 levels? Or should I split the data set and run 3 different analysis and then compare the results between them. The problem here is that the 3 periods differ in length and perhaps one of the periods is too short...

If this is not possible, Perhaps, I solution would be to test whether the co-occurrence of 2 species differs depending on the abundance of a 3rd species (abundance index should be entered in the DM as covariate)...

Thank you very much

Diego

[jhines]

You can build a model where detection is constant within a sub-period, but differs among sub-periods without using covariates. To do that, just use a different column in the design matrix for the 3 sub-periods. For example, if there were a total of 9 surveys and you wanted to define the first 3 surveys as sub-period 1, the 2nd 3 surveys as sub-period 2, and the last 3 surveys as sub-period 3, the design matrix would be:

-,b1,b2,b3
p1 1 0 0
p2 1 0 0
p3 1 0 0
p4 0 1 0
p5 0 1 0
p6 0 1 0
p7 0 0 1
p8 0 0 1
p9 0 0 1

You might compare the AIC from this model versus the AIC of a model where detection is constant or a model where detection is different for each survey to address your question about comparing detection probabilities.

You could do something similar with colonization or extinction, or with occupancy (if you use one of the alternate parameterizations).

Jim

[Diego.Pavon]

Thank you very much Jim!

I appreciate your comment

D