EDIT: Since researching the topic more, I now realize that I cannot simply exponentiate the log odds standard error to get standard error for the odds ratio. I think I will probably just use the log odds confidence intervals given and work with those instead.
Hi all-
I have seen several postings related to this but wanted to confirm my understanding of how to calculate the odds ratio and SE for an interaction term between a binary covariate and a continuous covariate in single season occupancy modeling (logit link).
Intercept = beta 1 (b1)
I have year as a covariate = 0 or 1 (b2);
A measure of human disturbance (HD) which is continuous (b3);
And an interaction term of year and HD (b4)
I believe that the odds ratio for HD in year 0 = e(b3)
and the odds ratio for HD in year 1 = e(b3+b4)
Because the odds ratio for HD in year 0 = e(b3), I used the SE for that covariate and did e(SE of b3) for the odds ratio SE.
The SE for the odds ratio of HD in year 1 is where I am less confident:
I printed the variance-covariance matrix for the betas and calculated the SE for the interaction in year 1:
e(sqrt(var(b3)+var(b4)+2*covar(b3,b4))).
Am I understanding this correctly?
Thanks so much!
- Molly