Probit Analysis

[Athinagoras]

Greetings,

My dependent variable is a binary variable (Yes/No). I chose to do a *Probit* analysis, in STATA.

Thus, I started by picking some ordered categorical variables from my dataset as my independent variables that I considered relevant.

Then,I run the Probit regression including dummy variables for each of the categories of my independent variables.

My question is:

I know from theory that, if the assumptions of probit regression are not met, then the coefficient estimates, standard errors and thus test statistics will not be correct.

I run a probit regression with a few independent variables and get values for coefficients, errors, etc. How does one proceed from here?

 

[ichbin]

There is nothing special in this regard about using a probit link function instead of, say, a logit link function. So all the goodness-of-fit procedures for logistic regression apply: http://en.wikipedia.org/wiki/Logistic_regression#Evaluating_goodness_of_fit.