• econometrics
• R
• linear probability model

The linear regression model is one of the most used tool not only in econometrics, but
in many other fields. The reason for this comes from its simplicity of estimation and
interpretation. However, there are situations in which this model does not appear
as the most natural choice. One of such situations is when the dependent variable is
binary. This thesis elaborates a method for estimation of the so-called linear proba-
bility model, implementing an algorithm to constrain the predicted probabilities to
be inside the (0,1) interval. This allows to extend the applicability of linear regres-
sion in a situation which is usually handled by logit or probit models that, however,
do not have a straightforward interpretation. The theoretical framework has been
implemented in an R package lpm.