An overview of multivariate analysis

Abstract

Multivariate statistical analysis is appropriate whenever several responses are measured on the same experimental unit. Univariate analysis applied separately to each response variable leads to incorrect tests of significance, since responses measured on the same experimental unit generally are correlated. Many multivariate techniques have univariate analogues and, therefore, can be easily understood and applied with only a general knowledge of matrix algebra. A wide variety of procedures exist which will meet the requirements of a wide range of experimental situations. High-speed computer programs for multivariate analysis have eliminated the tedious computations once required.