Is Item Imputation Always Better? An Investigation of Wave-Missing Data in Growth Models. Issue 4 (4th July 2021)

• Record Type: Journal Article
• Authors: Vera, Juan Diego; Enders, Craig K.
• Is Part Of: Structural equation modeling. Volume 28:Issue 4(2021)
• Page Start: 506
• Page End: 517
• Publication Date: 2021-07-04
• Subjects: Longitudinal studies -- missing data -- multiple imputation -- attrition -- questionnaires; Multivariate analysis -- Periodicals; Social sciences -- Statistical methods -- Periodicals; 519.535
• DOI: 10.1080/10705511.2020.1850289
• Languages: English
• ISSNs: 1070-5511

Abstract

ABSTRACT: Questionnaire data present challenges, as a missing item of a multi-item scale would lead to a total missing scale. A researcher applying multiple imputation to an incomplete multi-item questionnaire can impute the incomplete items prior to computing scale scores or impute the scale score entirely. Methodologist have favored item-level imputation because it greatly enhances precision in comparison to scale-level imputation; however, this benefit in precision might not translate into longitudinal data studies where entire questionnaire batteries are missing. We investigated the performance of item- and scale-level imputation model and found that item-level imputation did not produce a precision advantage in estimating any of the growth model parameters and scale-level showed better precision in estimating the slope variance parameter than item-level imputation.