A comparison of full information maximum likelihood and multiple imputation in structural equation modeling with missing data.

Abstract

This article compares two missing data procedures, full information maximum likelihood (FIML) and multiple imputation (MI), to investigate their relative performances in relation to the results from analyses of the original complete data or the hypothetical data available before missingness occurred. By expressing the FIML estimator as a special MI estimator, we predicted the expected patterns of discrepancy between the two estimators. Via Monte Carlo simulation studies where we have access to the original complete data, we compare the performance of FIML and MI estimators to that of the complete data maximum likelihood (ML) estimator under a wide range of conditions, including differences in sample size, percent of missingness, and degrees of model misfit. Our study confirmed well-known knowledge that the two estimators tend to yield essentially equivalent results to each other and to those from analysis of complete data when the postulated model is correctly specified. However, some noteworthy patterns of discrepancies were found between the FIML and MI estimators when the hypothesized model does not hold exactly in the population: MI-based parameter estimates, comparative fit index (CFI), and the Tucker Lewis index (TLI) tend to be closer to the counterparts of the complete data ML estimates, whereas FIML-based chi-squares and root mean square error of approximation (RMSEA) tend to be closer to the counterparts of the complete data ML estimates. We explained the observed patterns of discrepancy between the two estimators as a function of the interplay between the parsimony and accuracy of the imputation model. We concluded by discussing practical and methodological implications and issues for further research.

The Clinical Sustainability Assessment Tool: measuring organizational capacity to promote sustainability in healthcare

“…Confirmatory factor analysis (CFA) is a powerful and appropriate tool for testing a hypothesized subscale structure in a measurement instrument [34,35]. All CFA models were estimated using robust full-information maximum likelihood to efficiently handle any issues with normality and missing data [36]. Initially, CFA was applied to the pilot CSAT (with 7 items per subscale) to identify poorly performing items and test our hypothesized sustainability domain structure.…”

Curiosity helps: Growth in need for cognition bidirectionally predicts future reduction in anxiety and depression symptoms across 10 years

“…Following recommended best practices, the within-person cross-lagged relations (γs) in the RI-CLPM and within-person coupling effects in the BLCS (δs) were fixed to equality across timepoints to minimize standard errors of the parameter estimates (McArdle, 2009;Zainal & Newman, 2021a). To manage missing data, we used full information maximum likelihood; a gold standard approach in SEM for our dataset that was missing at random (Lee & Shi, 2021).…”

Substitute Behaviors following Residential Substance Use Treatment in the Western Cape, South Africa

“…As such, there is no evidence that both samples differ concerning the predictors gender, relationship status, living with a partner that uses substances, primary substance(s), and the covariate age. Full information maximum likelihood procedures, argued to yield equivalent results to multiple imputation, were used for missing values in the remaining analyses [ 56 ]. Three post-treatment outcome categories were constructed.…”