The Performance of the Full Information Maximum Likelihood Estimator in Multiple Regression Models with Missing Data

By Craig K. Enders. 2001

800 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/08f2bf7106a745acdd090a2f320ba951e2e6e60b

How researchers see this paper

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

In Statistics and Probability

Citations in the corpus

Motivating Employees of the Public Sector: Does Public Service Motivation Matter? (2010, background): “Hence, FIML provides estimators that have less bias and are less sensitive to sample variability (Enders 2001 ).”
Misplaced Trust? Exploring the Structure of the E-Government-Citizen Trust Relationship (2011, methodology): “Importantly, although OLS estimators should approach FIML (and ML) estimators for large samples under conditions of multi-variate normality (Bollen 1989), the FIML procedure has been shown to be more efﬁcient for analyzing data with missing observations (Enders 2001; Enders & Bandalos 2001).”
Missing Data in Educational Research: A Review of Reporting Practices and Suggestions for Improvement (2004, methodology): “However, there is evidence to suggest that normality violations minimally affect parameter estimate bias (Enders, 2001a; Graham & Schafer, 1999).”
The impact of kindergarten learning-related skills on academic trajectories at the end of elementary school (2006, background): “This approach produces parameter estimates that are less biased than listwise deletion, pairwise deletion or mean imputation (Enders, 2001; Wothke, 2000).”
Female and male antisocial trajectories: From childhood origins to adult outcomes (2008, methodology): “FIML is a widely accepted technique for dealing with missing data (Arbucle, 1996; Enders, 2001; Raykov, 2005).”
Biased attention and dysphoria: Manipulating selective attention reduces subsequent depressive symptoms (2010, methodology): “FIML has been shown to be superior to removing cases with missing data (ie, listwise deletion) for a several statistical techniques (Enders, 2001 ; Enders & Bandalos, 2001 ).”
Using an EM Covariance Matrix to Estimate Structural Equation Models With Missing Data: Choosing an Adjusted Sample Size to Improve the Accuracy of Inferences (2004, background): “…deletion, ML and MI parameter estimates exhibit less bias and greater efficiency in many circumstances (e.g., Arbuckle, 1996; Collins, Schafer, & Kam, 2001; Enders, 2001a, 2001b; Enders & Bandalos, 2001; Graham, Hofer, & MacKinnon, 1996; Graham & Schafer, 1999; Schafer, 1997; Wothke, 2000).”
Relating effortful control, executive function, and false belief understanding to emerging math and literacy ability in kindergarten. (2007, methodology): “FIML directly estimates model parameters and standard errors using all available data and has been shown to provide more accurate estimates of regression coefficients and variance accounted for than do listwise deletion or mean replacement missing data strategies (Enders, 2001).”
Physical Activity, Self-Regulation, and Early Academic Achievement in Preschool Children (2014, methodology): “Full information maximum likelihood estimation utilizes all available information and has been shown to produce less biased estimates than listwise deletion (Acock, 2012; Enders, 2001).”
Relationship Between Vicarious Traumatization and Turnover Among Child Welfare Professionals (2015, methodology): “Missing data were determined on all key measures and the full information maximum likelihood (FIML) estimator implemented in MPLUS was utilized (Enders, 2001; Raykov, 2005).”
Construct Validation of the Behavior and Instructional Management Scale. (2010, methodology): “As the amount of missingness increases or data are MAR or MNAR, traditional missing data methods (e.g., listwise deletion, pairwise deletion, simple imputation, mean imputation, etc.) often produce biased parameter estimates and reduced statistical power (Enders, 2001).”
Preschool Children’s Prosocial Behavior: The Role of Mother–Child, Father–Child and Teacher–Child Relationships (2016, methodology): “Next, using structural equation modeling (SEM) procedures, we tested our hypothesis by fitting a series of structural equations models to the data, applying full information maximum likelihood estimation (Enders 2001).”
Teacher mindset is associated with development of students' growth mindset (2021, methodology): “…FIML has been shown to yield less biased parameter estimates and less sampling variability compared to listwise deletion and mean imputation (Enders, 2001), which afforded us a robust method to retain cases with missing data only on continuous measures of interest (e.g., student growth…”