The theoretical properties of maximum likelihood estimators and their use in overcoming shortcomings of the classical linear model. Computer algorithms are developed and used to compute maximum-likelihood estimators for logit, probit, tobit, sample-selectivity, and failure time models.