untransformed beta values in two-species model output

[ceosborne]

Hi, I'm analyzing secretive marsh bird data. I ran individual single-species and Royle-Nichols models for each of the 5 species included in the surveys. When I ran Virginia rail and Sora together in the two-species model, the output values are listed as follows:

Untransformed Estimates of coefficients for covariates (Beta's)
==============================================================================
estimate std.error
A1 ccupancy psiA 1.557009 (0.381411)
A2 ccupancy psiB -1.151646 (0.350525)
A3 ccupancy psiBa -15.767630 (32168391426.155121)
B1 :detection pA[1] 0.796036 (0.236718)
B2 :detection pB[1] -10.910719 (31702492412.185669)
B3 :detection rA[1] 1.962942 (0.537150)
B4 :detection rB[1] 1.047778 (0.463629)
B5 :detection rBa[1] -0.083154 (1.013258)

How can I transform these values to be bound between 0 and 1? I am very new to this, but from what I have read, r must not be bound in the model or lamda will be undefined. And the logit link cannot be used to transform the results because this is not a binary model. Whst can I do? I'm don't know how to interpret these results, with the exception of the huge SEs up there!

thank you!
Carrie

[jhines]

Hi Carrie,

PRESENCE should list the transformed values after the 'Beta' estimates. If not, please send me the input(pao) and output(pa2) files off-list.

The link function for the parameters in this model is:

Code: Select all

           exp(A1)
psiA = ---------------------
          (1+exp(A1)


The same function would apply to all parameters, including r. r is the probability of detection, given both species are present.

If you use the other parameterization, there are parameters which are not bound between zero and one. For example, lambda (delta in book), which is the species interaction factor can be greater than one. In this case, PRESENCE uses a log transformation (instead of logit):


lambda = exp(A1)

Cheers,

Jim

[ceosborne]

Thank you! I sent the files to you.