An International Marketing Application of Outlier Analysis for Structural Equations: A Methodological Note

Abstract

Structural equation modeling with latent variables is being used more frequently in international marketing research. However, the authors argue that it is hazardous to conduct cross-national marketing research without evaluating the potential influential effects of multivariate outliers, which are observations distinct from the majority of cases. Because the presence of outliers in the data can significantly bias a study's findings, this is an important issue in international research. To improve upon current practice, the authors recommend using a two-step approach for detecting and analyzing multivariate outliers in structural equation models. The first step is to detect outliers using three techniques: Bollen's a _ii (1987), Mahalanobis Distance, and the Observed Covariance Ratio, a new technique developed by the authors. The second step is to determine whether outliers unduly influence study findings. This is accomplished by estimating statistical models with and without outliers and comparing results. The authors demonstrate the two-step approach using data from a previous international marketing study. Several outliers were found to influence model fit, R ² , and the size and direction of parameter estimates. The study highlights the importance of multivariate outlier analysis to international researchers.