Collapsing Sparse Responses: Examination of Confirmatory Factor Analysis Models Using Likert Scale Data

• In Event: Saturday Roundtable Session 3:05 pm
  In Roundtable Session: Regression Modeling for Educational Research (Table 21)

Abstract

Applied researchers often encounter situations where few respondents endorsed certain item response categories resulting in sparse data. Collapsing categories is a common method to deal with such data issues. The current study was conducted through a simulation of a commonly used confirmatory factor analysis model with sparse data. Afterwards, data was recoded to collapse responses for items with sparse data. The results indicated that collapsing categories increase the performance of model fit for the conditions under the correctly identified model. However, collapsing responses in misspecified models may increase the chance of selecting an inappropriate model as well. The results are of interest to researchers to inform practices for handling sparse data in CFA.

Authors

• Jin Liu, University of Texas - Arlington

• Yu Bao, James Madison University

• Christine DiStefano, University of South Carolina