The Relative Performance of Full Information Maximum Likelihood Estimation for Missing Data in Structural Equation Models

By Craig K. Enders, D. Bandalos. 2001

4,653 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Management Science and Operations Research, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/cb410defa1263e27b60b7ab86ec72fca3f75027b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

1. Victoria Savalei
2. Alexandre J. S. Morin
3. Peter M. Bentler
4. John P. Meyer
5. Daniel W. Russell
6. Jörg Matthes
7. Michael J. Kieffer
8. Mikki Hoang Phan

In Management Science and Operations Research

1. Victoria Savalei
2. Alexandre J. S. Morin
3. Herbert W. Marsh
4. Peter M. Bentler
5. Kristopher J Preacher
6. Benjamin Nagengast
7. Thomas A. Schmitt
8. Ross Larsen

In Statistics and Probability

1. Victoria Savalei
2. Peter M. Bentler
3. Herbert W. Marsh
4. Alexandre J. S. Morin
5. Daniel W. Russell
6. Carl F. Falk
7. Mark L. Davison
8. Jeffrey D. Long

Citations in the corpus

• The Development and Validation of the Game User Experience Satisfaction Scale (GUESS) (2016, positive): “FIML has been demonstrated to generate unbiased parameter estimates and standard errors, and valid model fit information when data are not missing completely at random and across different normality conditions (Dong & Peng, 2013; Enders, 2001; Enders & Bandalos, 2001; Hallgren & Witkiewitz, 2013).”
• Early Childhood Teachers’ Psychological Well-Being: Exploring Potential Predictors of Depression, Stress, and Emotional Exhaustion (2018, methodology): “Full information maximum likelihood estimation preserves all available data to produce unbiased estimates under conditions of both missing completely at random and missing at random (Enders & Bandalos, 2001).”
• The effects of early socialization experiences on content mastery and outcomes: A mediational approach (2006, methodology): “We also used full-information maximum likelihood (FIML) estimation to estimate model parameters from raw data with missing values (Enders & Bandalos, 2001).”
• The Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5): Development and Initial Psychometric Evaluation. (2015, methodology): “Missing data were addressed using full information maximum likelihood (FIML), and acceptable covariance coverage was found (82.7% to 95.2% for trauma measures, 29.0% to 97.3% for other measures of psychopathology; Enders & Bandalos, 2001).”
• In Search of Underlying Dimensions: The Use (and Abuse) of Factor Analysis in Personality and Social Psychology Bulletin (2002, methodology): “Monte Carlo studies have indicated that this procedure provides a more accurate estimate of model parameters than methods such as the exclusion of cases with missing data when missingness is related to one or more of the variables included in the analysis (see, e.g., Enders & Bandalos, 2001).”
• Effects of Missing Data Methods in Structural Equation Modeling With Nonnormal Longitudinal Data (2009, methodology): “Although the method has desirable qualities, such as ease of implementation, it can waste a great deal of incomplete data, resulting in inefficient parameter estimates ( Enders & Bandalos, 2001 ).”
• Minority Stress Among Transgender Adolescents: The Role of Peer Victimization, School Belonging, and Ethnicity (2018, positive): “Full information maximum likelihood (FIML) estimation was employed to handle missing data as it has been found to be a less biased and more powerful approach when compared to other options (Enders and Bandalos 2001; Tabachnick and Fidell 2013).”
• Is the ML Chi-Square Ever Robust to Nonnormality? A Cautionary Note With Missing Data (2008, result): “This is in line with the previous findings reporting good performance of this statistic (Arbuckle, 1996; Enders & Bandalos, 2001).”
• A critical review of structural equation modeling applications in construction research (2015, methodology): “not a good option when there are more than 5% missing values per indicator as this decreases the variability of data [26]; (2) a returned questionnaire with more than 15% missing values should be treated as an invalid response [26]; and (3) the full information maximum likelihood (FIML) method is mo”
• Converging Trajectories (2011, result): “…that has been shown in Monte Carlo studies to yield equal or superior results to other approaches to missing data in SEM applications (e.g., Enders & Bandalos, 2001); each SEM model was then fitted to the estimated variance-covariance matrix, consistent with the recommended default…”
• (Why) Should We Use SEM? Pros and Cons of Structural Equation Modeling (2003, methodology): “The SEM framework has been enriched with the newest procedures to handle missing data (Enders, 2001; Enders & Bandalos, 2001) and was extended to moderator models.”