odd ratio models
Message posted by Garrett Sonnier on January 14, 2000 at 12:00 AM (ET)

I have a question concerning a model that I'm currently working on The dependent variable in my model is the log odds ratio (ln(p/(1-p)). I use OLS to estimate the model using two continuous variables and a dummy variable. The two continuous variables are transformed to logarithmic scale. I want to know the % change in p given the % change in a particular independent variable. How do I interpret the OLS coefficiennts where the DV in the model is log(p/1-p). Furthermore, is the Halvorsen Palmquist approach correct to deduce the % impact associated with a change in dummy status from zero to one?