Ignoring Multilevel Data Structure in Confirmatory Factor Analysis With Ordinal Items

• In Event: 34.079 - Roundtable Session 14
  In Roundtable Session: 34.079-11 - Ordinal and Binary Data: From Differential Item Functioning to Classification and Regression Tree

Abstract

This study compared multilevel CFA, single-level CFA with adjusted standard error (SE) and that with normal SE in analyzing the factor structure of ordinal item responses from clustered samples. Data were simulated with different factorial structure, ICC, and sample size. The single-level CFA with adjusted SE provided good fit to the two-level data. The single-level CFA with normal SE provided good fit only when ICC was low. When the single-level model was used to estimate the data with clustering, the factor loading was more robust than the factor correlation especially when the within-level and between-level factor loading were the same. The factor loading was positively (negatively) biased when the between-level loading was higher (lower) than the within-level loading.

Authors

• Li Zhou, University of Pittsburgh

• Feifei Ye, University of Pittsburgh