Publication Date



By determining whether independent or dependent variables, or both, are multivariate and specifying the scale of measurement of the variables an appropriate multivariate statistical procedure can be identified. Reference are given for locating additional information on these procedures.

For many years Tatsuoka and Tiedeman's (1963) cross partition in Gages' Handbook of Research in Teaching has been a useful tool for identifying an appropriate statistic for a particular research question. Since its publication many new statistical procedures have been developed, particularly for use with multivariate designs. The purpose of this article is to incorporate the new procedures and those in Tatsuoka and Tiedeman in a new set of cross partitions focusing only on multivariate cases.

A study is multivariate if there is more than one independent variable and/or more than one dependent variable. Some of the techniques used with multivariate statistics such as multiple regression are well known. Other multivariate techniques are available for other combinations of variables, some of which are not so familiar.

Table I lists statistical procedures appropriate when the independent variable is multivariate and the dependent variable is univariate. When the dependent variable is multivariate and the independent variable is univariate, Table II is appropriate. Finally, where both the independent and dependent variables are multivariate Table Ill should be consulted. (Factor analysis, a commonly used multivariate statistical procedure, is not included in any of these table, since it is used when the variables are not classified as independent or dependent.) After locating the proper table one must identify the scale of measurement of the variables (Stevens, 1946).

Included in

Education Commons