I am interested in the effects of gender on a set of variables. I have run a number of group mean t tests (with gender as the group variable), and have confirmed that the means for men and women across the different variables are different at statistically significant levels.
This analysis may in itself be enough to write a paper, as the variables are interesting and the differences are backed by theory. However, would it be appropriate to use the dummy variable gender as an outcome variable in a logistic regression in order to see which of the dependent variables are most predictive? My guess is that this is not possible (theoretically), since obviously gender should be understood as an independent rather than a dependent variable. I am hoping that someone might confirm this for me - or is a logistic regression on gender possible?
Yes, from this description it does not seem appropriate since income would not influence gender which is set. Perhaps you may want to better describe why you would want to run these follow-up analyses, what are you trying to get at in particular with the logistic regression?
Thanks for the response. I think you have understood my purpose: the variable I am interested in is gender, but as a predictor of a number of variables (income and several others). The other variables are also correlated, and so if I did want to do several regressions on the different variables, I would have to control for the other variables as well (income and the others), and this seems even less appropriate - just playing with the numbers with very little theory.
Anyways, thanks again. --Also, I think now that I'm posting in the wrong forum. 'Statistical research' I'm guessing now is more directed towards people doing research about statistics, while I am using them in social science research. Cheers.