The logit model can be derived by assuming an underlying model for a continuous variable which is then dichotomized. While this rationale is not essential, it can be helpful in understanding several aspects of the model:
How the model might have arisen in the real world.
Where standardized coefficients come from.
How the logit model is related to the probit and complementary log-log models.
How the model can be generalized to ordinal dependent variables.
How unobserved heterogeneity can affect logit coefficients.
In this section, we’ll examine the first two points, leaving the other three for later.
Here’s the idea. Suppose we have a dependent variable z that is measured on a continuous scale. ...