The problems of two way analysis of variance designs with unequal and disproportionate cell sizes are discussed. A variety of solutions are discussed and a new solution is presented. (JKS)

Problems associated with the use of gain scores, analysis of covariance, multicollinearity, part and partial correlation, and the lack of rectilinearity in regression are discussed. Particular attention is paid to the misuse of statistical techniques. (JKS)

With even the simplest bivariate regression, least-squares solutions are inappropriate unless one assumes a priori that reciprocal effects are absent, or at least implausible. While this discussion is limited to bivariate regression, the issues apply equally to multivariate regression, including stepwise regression. (Author/CTM)