Random Contamination and Select Response Styles Affecting Measures of Fit and Reliability in Factor Analysis

• In Event: 32.080 - Measurement and Validity Issues in Survey Research

Abstract

This research examines the effects of nonattending response pattern contamination and select response style patterns on measures of model fit (CFI) and internal reliability (Cronbach’s α). A simulation study examines the effects resulting from percentage of contamination, number of manifest items measured and sample size. Initial results indicate that sample size very mildly affects CFI but does not influence α. Percent contamination decreases both CFI and α in a nearly linear fashion over a limited range of contamination. Finally, whereas an increase in the number of manifest items increases resilience to random contamination for α, the opposite was observed for CFI. An increase in the number of manifest items resulted in larger decreases in CFI. Implications are briefly discussed.

Authors

• Allen G. Harbaugh, Boston University

• Min Liu, University of Hawaii - Manoa