8. Outliers, Leverage Observations, and Influential Cases in Factor Analysis: Using Robust Procedures to Minimize Their Effect

摘要: Parallel to the development in regression diagnosis, this paper defines good and bad leverage observations factor analysis. Outliers are that deviate from model, not center of data cloud. The effects each kind outlying on normal distribution-based maximum likelihood estimator associated ratio statistic studied through distinction between outliers also clarifies roles three robust procedures based different Mahalanobis distances. All designed minimize effect certain observations. Only procedure with a residual-based distance properly controls outliers. Empirical results illustrate strength or weakness support those obtained relevance general structural equation models is discussed formulas provided.