The Impact of Nonattending Response Style on Statistical Analyses With Likert-Type Scales

• In Event: Data Consideration in Survey Research

Abstract

This research examines the contamination effects of non-attending response style (NARS) on measures of model fit. A simulation study examines the effects resulting from percent contamination, number of reverse-coded items, and magnitude of factor loadings. Initial results indicate that the influence of percent contamination on model fit (measured via the uncensored Comparative Fit Index) follows a U-shaped non-linear trend with worst impact at the 50% contamination mark. General results suggest more reverse-coded items increases detrimental effects on model fit. Furthermore, weaker factor loading strength seems to result in less impact on model fit. Curious results were observed for the scenario with exactly half of the items being reverse-coded, warranting further study. Results clearly indicate that NARS requires further investigation.

Author

• Allen G. Harbaugh-Schattenkirk, Longwood University