Analyzing distances

Gower and Krzanowski (1999) described the analysis of a multivariate dataset that violated the assumptions of normality for multivariate analysis of variance. They developed a method comprising two aspects: a graphical representation of the points in a fewer number of dimensions, known as principal coordinate analysis; and a technique similar to MANOVA except that it was based on partitioning the distances between subjects, rather than sums of squares, and did not assume that the data followed any particular distribution. This article summarizes both aspects of their analysis and describes the Stata pco and aod commands, which perform principal coordinate analysis and analysis of distance. Copyright 2004 by StataCorp LP.