Applying Hultivariate Analyses Inappropriately

Abstract

It bccones easier to use nultivariatc analyses than before when personal computers and statistical software packages were not available. Social and behavioral sciences also benefited by then which stinulated various kinds of researches in these fields. But the stimulation induced negative effect on these fields as well, i.e. inappropriate application of multivariate analyses. They can be classified into two types. One is not directly related with the analysis but caused by the earlier stages which precede the analysis. The other is in the analysis itself. Although personal computers and statistical software packages have accclcratcd the second type of the inappropriate applications of nultivariate analyses, there are many kinds of inappropriate applications of nultivariate analyses which have no relationship with personal computers and statistical software packages but caused by the researchers themselves. In the present paper we focus our attention onto the second type of inappropriate applications of oultivariate analyses. Typical examples of inappropriate applications of nultiple regression analyses, factor analyses, cluster analyses, and multidimensional scaling analyses are presented and discussed. Problems which cause inappropriate applications of statistical or quantitative analysis procedures are talked over, and some neasures to overcome then are proposed. © 1993, Japanese Association For Mathematical Sociology. All rights reserved.