Use of Two-Block Partial Least-Squares to Study Covariation in Shape

Abstract.—The relatively new two-block partial least-squares method for analyzing the covariance between two sets of variables is described and contrasted with the well-known method of canonical correlation analysis. Their statistical properties, types of answers, and visualization techniques are discussed. Examples are given to show its usefulness in comparing two sets of variables—especially when one or both of the sets of variables are shape variables from a geometric morphometric study. [Canonical correlation; covariance; morphometrics; Mus musculus domesticus; partial least squares; Plethodon; visualization.] In this article we describe the application to morphometrics of two-block partial least-squares (2B-PLS) analysis (Sampson et al., 1989; Streissguth et al., 1993), a relatively new statistical method that can be used to explore patterns of covariation between two sets of variables. This method differs from regression analysis in that the two sets o