Marginal effects of discrete variables in logistic regression

[Yashu Kalera]

Hi everyone
I am conducting a logistic regression analysis, and am using discrete independent variables. Specifically, my variables include year (=0,1,2), Group (=0 for control, =1 for treatment), and other variables such as gender (=1 for male, =2 for female). Could somebody please clarify the difference between the commands

(1) margins Group year, dydx(sexo) pwcompare(cimargins effects) predict(xb) {are the reported estimates in odds ratios with this command? How should I interpret the result of this command?}
(2) contrast r.year r.Group r.sexo r.year#r.Group#r.sexo, or {I have chosen the odds ratios option with this, but am again struggling with the interpretation.}

(sexo is the gender variable)

Thank you

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. If you're having trouble with odds ratio, move to logit and work in predicted probabilities.

In (1), you're asking for xb which I don't think is an odds ratio. It looks like you're asking for dydx on xb which should look a lot like b. This is documented in logistic postestimation. Read the documentation.

I personally prefer predicted probabilities. I often explicitly set the values for the iv's (e.g., at(sex=(0 1)) )and then look at the difference in predicted probabilities. I know this is not the most efficient way to do it in Stata, but I can understand it.