Binary dependent variable regression: probit or logit?

[jennystats]

Hi guys,

I'm trying to run a regression in Eviews 6 Student Version where my dependent variable is a binary variable, and most of the explanatory variables are binary as well except for one or two.

In class I learned that we can just us OLS on this and it is called a linear probability model, but that sometimes it makes more sense to use probit or logit regressions if it does not make sense that probability should be linear in X (if sometimes you get probabilities <0 or >1).

How do I know when to use probit or logit as opposed to just OLS? And how would I run one of those regressions?

Any help appreciated! Thank you