GUIDELINES FOR CHOOSING BETWEEN ACOV AND PLS-PM FOR ORDINAL DATA

Abstract

The accuracy on parameter recovery is compared between Structure Covariance Analysis (ACOV) and Partial Least Squares Path Modeling (PLS-PM), with simulated ordinals data with 5 points, in a simple model. An experimental design is used, controlling the estimation method, sample size, skewness level and model specification. Mean absolute differences are used to assess accuracy for the structural model. ACOV provided more accurate estimates of the structural parameters than PLS-PM in different experimental conditions. With a small sample size, both techniques are equally accurate. Using ACOV against PLS -PM is suggested. PLS choosing ACOV instead based on the use of a small sample size is not recommended.