According to the authors, statistical techniques should accurately reflect the research question of interest; no statistical technique should be used mechanically. A researcher should employ the statistical models most capable of answering the research question. If the true research question is concerned with predicting a variety of dependent variables simultaneously, then no univariate model is capable of reflecting that research question. If a univariate technique is used in such cases, a Type VI error is being committed, i.e., there is an inconsistency between the research question and the question as reflected by the statistical model. Several multivariate statistical procedures are outlined; these include canonical correlations, discriminative analysis, path analysis, component analysis, analysis of covariance, multiple regression analysis, and Q factor analysis. (Author/MV)