ECON 104 Lecture Notes - Lecture 39: Function Problem, Likelihood Function, Stata
Document Summary
Logit regression is like probit but probit uses nonlinear cdf while logit uses logistic cdf (functional form of cdf changes) Logit is more beneficial because it has an analytical expression. How do you write prob y = 1 given x. Can plug in betas and x"s and then you get the f thing. Logit is computationally faster (just analytically computable value), but probit is more to do with a table. Since computers are much faster than probit and logit was first developped, speed doesn"t really matter. Both have very similarly estimated curves and y hat estimates. Empirical results won"t hinge on logit/ probit choice, both tend to be used in practice. Many people who do research with binary dependent variable, they will report it for probit and logit regression. When you make a mistake, you will recognise that you are making a mistake. Instead of probit you say logit in stata.