« Term Paper Research | Main | Extensions of logit/probit »

November 14, 2006

logit/probit

Calling this the "linear probability model" makes it sound better than it really is. The regression model is simply not the best option for categorical data.

PB140001a.jpg

The probit model is designed for this situation.

PB140004a.jpg

PB140004b.jpg

The logit model is another excellent choice.

PB140006a.jpg

The odds ratio

PB140008a.jpg

leads to the log odds ratio.

The likelihood ratio test is our bread and butter test for logit/probit models.

PB140010a.jpg

Posted by bparke at November 14, 2006 09:41 PM

Comments