EVALUATING FIML AND MULTIPLE IMPUTATION IN JOINT ORDINAL-CONTINUOUS MEASUREMENT MODELS WITH MISSING DATA

Abstract

Missing data is a common occurrence in confirmatory factor analysis (CFA). Much work has evaluated the performance of different techniques when all observed variables were either continuous or ordinal. However, few have investigated these techniques when observed variables are a mix of continuous and ordinal variables. This study investigated the performance of five approaches to handling missing data in these models, a joint ordinal-continuous full information maximum likelihood (JOC-FIML) approach and four multiple imputation methods — two imputation algorithms (fully conditional specification and expectation-maximization with bootstrapping) combined with two estimators (robust maximum likelihood and weighted least squares with mean and variance adjustment). In a Monte-Carlo simulation, the JOC-FIML approach performed the best in estimating factor loadings, standard errors, and model fit in almost all conditions. We recommend JOC-FIML across most conditions, except when certain ordinal categories have extremely low frequencies as it is more likely not to converge and produce inaccurate model fit statistics when low frequencies are present. If the sample is large, fully conditional specification combined with weighted-least-squares is recommended when low frequencies are present or when the FIML approach is not feasible (e.g., when variables that predict missingness are not of interest to the analysis).