NLOGIT implements full-information maximum-likelihood or simulated-likelihood estimators for a wide range of choice models. Among them are: • Unrestricted and scaled
multinomial logit models, including heteroscedastic variants. •
Mixed logit (a.k.a. random-parameters logit) with flexible distributional assumptions. • Random-regret logit and “willingness-to-pay-space’’ parameterisations. •
Nested logit,
heteroscedastic extreme value and error-components logit. •
Multinomial probit with simulation-based integration. •
Latent class models for unobserved taste heterogeneity. ==Data analysis and simulation==