ECON 104 Lecture Notes - Lecture 38: Linear Probability Model, Nonlinear Regression
Document Summary
Black variable = 1 if applicant is black, 0 otherwise. How does probability of denial get impacted by black or not. Would put 0 for binary variable and compute predicted variable as ___ Read off the coefficient in front of dummy and say that the difference between races in terms of denial probability is 17. 7% Inference is the same, use robust standard errors. Disadvntages = present in linear estimation, change in predicted prob for a given change in x is the same for all values of x. Regardless of where you start, it will give you same change in y. Linear prob model commonly predicts probability to be < 0 or > 1. Can solve these disadvantages by using a nonlinear probability model - probit and logit regressions. In linear models, prob depends linearly on x. Instead we want prob to increasing in x for beta 1 > 0, and p between 0 and 1 for all x.
