When MICE and FIML Disagree: Investigating Inconsistencies in Missing Data Analysis 15 views

Missing data are pervasive and pose a significant challenge in most psychological studies. Various techniques have been developed to handle missing values. As a family of multiple imputation algorithms, Multivariate Imputation by Chained Equations (MICE) is widely applied due to its flexibility and practical usefulness in dealing with diverse types of data and models. However, recent empirical studies showed that different MICE imputation algorithms and full information maximum likelihood (FIML) estimation may lead to inconsistent statistical conclusions, indicating that the performance of MICE algorithms and the comparison between MICE and FIML remain underexplored. Therefore, this study systematically compares the performance of FIML and commonly used MICE algorithms, including predictive mean matching (PMM), Bayesian linear regression (norm), and classification and regression trees (CART) for linear regression models, by manipulating sample size, effect size, number of auxiliary variables, missing data mechanism, and missing data rate. Our results show that while MICE is a powerful tool, some default settings of MICE algorithms can lead to biased model estimations and should be used with caution.

MA (Master of Arts)

English

All rights reserved by the author (no additional license for public reuse)

2026-02-05