Abstract
Since the passage of the major civil rights legislation of the sixties, the courts have been entertaining ever more sophisticated statistical arguments in discrimination cases. A review of this development clarifies the numerous conceptual and technical difficulties inherent in the regression approach to analysis of covariance. A nonparametric analysis based on matching is supported as more responsive to the fundamental question involved: whether individuals who are similar with respect to all pertinent variables are treated differently depending on their membership in a particular class.
Get full access to this article
View all access options for this article.
