Discrete Choice 1.0 for GAUSS 5.0

Discrete Choice is a package for the fitting of a variety Discrete Choice 1.0 of models with categorical dependent variables. These models are particularly useful for researchers in the social, behavioral, and biomedical sciences, as well as , public choice, Models Include: education, and marketing. z Nested model Output for these models includes full information maximum z Conditional logit model likelihood estimates with either standard and quasi-maximum likelihood inference. In addition, estimates of marginal effects z Multinomial logit model are computed either as partials of the probabilities with respect z Adjacent category to the means of the exogenous variables or optionally as the multinomial logit model average partials of the probabilities with respect to the z Stereotype multinomial logit exogenous variables. model • Nested logit model • Stereotype multinomial logit model z Poisson and negative Is derived from the assumption that The coefficients of the regression in , left or residuals have a generalized each category are linear functions of right truncated, left or right extreme value distribution and a reference regression. censored, or zero-inflated allows for a general pattern of • Poisson and negative binomial dependence among the responses models regression, left or right truncated, thus avoiding the IIA problem, i.e., left or right censored, or zero- the “independence of irrelevant z Logit, probit models inflated models alternatives.” Estimates model with Poisson or z , probit models • Conditional logit model negative binomial distributed Includes both variables that are dependent variable. This includes attributes of the responses as well censored models—the dependent as, optionally, exogenous variables variable is not observed but that are properties of cases. independent variables are available— Requirements: • Multinomial logit model and truncated models where not even GAUSS Mathematical & Statistical Qualitative responses are each the independent variables are System (GAUSS) Version 5.0.30+ or modeled with a separate set of observed. Also, a zero-inflated the GAUSS Engine/GAUSS Engine regression coefficients. Poisson or negative binomial model Pro/GAUSS Engine for Workgroups/ • Adjacent category multinomial can be estimated where the probability of the zero category is a GAUSS Enterprise Engine v5.0.30+ logit model mixture of a negative binomial The log-odds of one category consistent probability and an excess Platforms: versus the next higher category is probability. The mixture coefficient Available for Windows, LINUX, and linear in the cutpoints and can be a function of independent UNIX: AIX4, Sun Sparc and HPUX11. explanatory variables. variables.

For More Information Contact: Aptech Systems Inc. 23804 SE Kent-Kangley Road Maple Valley, WA 98038 USA Phone: (425) 432-7855 z FAX: (425) 432-7832 Email: [email protected] z URL: www.Aptech.com