Reconstituting parameter estimate w/ normalized covariates

[qrobinson]

Hi everyone,

I have a question concerning how to reconstitute a parameter estimate (occupancy) from beta coefficients of a normalized covariate.

I understand the mechanics of how to reconstitute an occupancy estimate for different values of a non-normalized covariate, given the logit-scale beta estimates provided by the output file. However, I'm wondering how to handle this procedure with a covariate that's been z-normalized prior to the analysis. I suspect that (on the logit scale) my covariate's beta coefficient represents the change in the response variable due to a 1-unit change in the covariate on the z-scale. Is this correct?

If so, does anyone have any suggestions regarding how to transform the various components of the equation back into biologically relevant units?

Thanks as always for your time and assistance!

Cheers,

Quinn

[cooch]

Hi everyone,

I have a question concerning how to reconstitute a parameter estimate (occupancy) from beta coefficients of a normalized covariate.

I understand the mechanics of how to reconstitute an occupancy estimate for different values of a non-normalized covariate, given the logit-scale beta estimates provided by the output file. However, I'm wondering how to handle this procedure with a covariate that's been z-normalized prior to the analysis. I suspect that (on the logit scale) my covariate's beta coefficient represents the change in the response variable due to a 1-unit change in the covariate on the z-scale. Is this correct?

If so, does anyone have any suggestions regarding how to transform the various components of the equation back into biologically relevant units?

Thanks as always for your time and assistance!

Cheers,

Quinn

Approach for linear models in general is described in Chapter 11 of the MARK book. See various pages, starting with 12-13.

[qrobinson]

Hi,

Thank you Dr. Cooch for the direction -- I should have caught that this was covered in the FM. Apologies.


Best,

Quinn