The analysis and prediction using binary discrete choice models

In this paper probabilistic models have been proposed for learning and analysis about discrete choice behaviour. The first problem discussed in the article above is the specification and estimation of parameters of three models, the linear probability model and the nonlinear logit and probit models of binary choice. Some of the difficulties associated with the interpretation of the linear probability model suggest the usefulness of the nonlinear probability models. If only one or a few observations exist for each decisionmaker, maximum likelihood estimation is possible for the two models, logit and probit. When enough observations are available one can estimate the parameters of probit and logit models also by generalized least squares method. Given the availability of efficient computer packages (LIMDEP, RATS, GAUSS, SST, TSP) there are little differences in the computational effort of the two estimation methods.