Multiple regression with a qualitative dependent variable

Abstract

This paper describes several methods by which a single set of scores for a qualitative (usually ordinal) dependent variable can be estimated simultaneously with the coefficients of the explanatory variables of a model. The canonical correlations and multiple discriminant analysis approaches are well known in the statistics literature. However, the paper goes on to show an iterative least-squares multiple regression technique can provide a useful approximation to these more general procedures. The techniques are illustrated with labor force participation and voter turnout examples.