Constrained Minimum Discrimination Information: A Unifying Tool for Modeling Spatial and Individual Choice Behavior

Abstract

Mathematical programming and statistical inference are combined in a constrained minimum discrimination information (MDI) method to provide a basis for a wide range of spatial and individual choice behavior problems. This approach offers an alternative to linear and loglinear regression estimation methods as well as probabilistic models of the logit and probit variety. Some logical and computational difficulties inherent in these approaches are resolved. Further, the approach leads endogenously to alternative hypotheses if the null hypothesis is rejected, and hence has implications for the interaction between research that is oriented toward theory construction and applied research that is empirically oriented.