The impact of nonnormality on full information maximum-likelihood estimation for structural equation models with missing data.

By Craig K. Enders. Psychological methods, 2001

720 citations recorded by the source corpus.

Subfields: Management Science and Operations Research, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/8e42765f9bad8253bd671101ab653ab99988da9d

How researchers see this paper

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

In Management Science and Operations Research

1. Victoria Savalei
2. Carl F. Falk
3. Alexandre J. S. Morin
4. Peter M. Bentler
5. Benjamin Nagengast
6. John Th⊘gersen
7. Herbert W. Marsh
8. O. Schilling

In Statistics and Probability

1. Victoria Savalei
2. Carl F. Falk
3. Peter M. Bentler
4. Mark L. Davison
5. Jeffrey D. Long
6. Tacksoo Shin
7. Craig K. Enders
8. Alexandre J. S. Morin

Citations in the corpus

• The Relationships Among Students' Future-Oriented Goals and Subgoals, Perceived Task Instrumentality, and Task-Oriented Self-Regulation Strategies in an Academic Environment (2008, background): “, Brickman & Miller, 2001; Greene et al., 2004; Miller & Brickman, 2004; Raynor, 1970, 1974; Vansteenkiste, Simons, Lens, Soenens, et al., 2004). For example, Miller and Brickman (2004) as well as Greene et al.”
• Effects of Missing Data Methods in Structural Equation Modeling With Nonnormal Longitudinal Data (2009, methodology): “Enders (2001a) reported that the standard errors and model fit statistic (chi-square statistic) of robust ML were a dramatic improvement over the results of standard ML.”
• The Influence of Childhood (2012, background): “Data were incomplete and deemed missing at random (Allison, 2003; Schafer & Graham, 2002), showing mild departures from normality, with absolute univariate skewness and kurtosis values slightly different from zero (Enders, 2001).”
• Performance of Estimators for Confirmatory Factor Analysis of Ordinal Variables with Missing Data (2020, methodology): “, Enders & Bandalos, 2001) and continuous form of nonnormality (Enders, 2001).”
• A Comparison of Maximum-Likelihood and Asymptotically Distribution-Free Methods of Treating Incomplete Nonnormal Data (2003, methodology): “Enders (2001) compared full information ML estimation with listwise deletion, pairwise deletion, mean imputation, and similar response pattern imputation with respect to parameter estimate bias and efficiency, confidence interval coverage, and rate of model rejection, in models with nonnormal…”
• Creating College Opportunity: School Counselors and Their Influence on Postsecondary Enrollment (2013, background): “…thereby allowing all cases (with at least one data point) to contribute to the maximum likelihood function.10 9 It is important to note that while the normal distribution plays an essential role in maximum likelihood estimation, FIML is quite robust in the presence of non-normal data (Enders 2001).”
• The Impact of Missing Data on Sample Reliability Estimates: Implications for Reliability Reporting Practices (2004, background): “These theoretical expectations have been supported by a number of recent empirical studies (e.g., Arbuckle, 1996; Collins et al., 2001; Enders, 2001a, 2001b; Enders & Bandalos, 2001; Graham et al., 1996; Schafer & Graham, 2002; Wothke, 2000).”
• Multiple Imputation Strategies for Multiple Group Structural Equation Models (2011, background): “Rather, the user specifies the set of variables that will be used during imputation (e.g., by listing the seven variables on the VAR subcommand of PROC MI), and the software selects the variables that are appropriate for a given missing data pattern.”
• A Statistically Justified Pairwise ML Method for Incomplete Nonnormal Data: A Comparison With Direct ML and Pairwise ADF (2005, methodology): “…than direct ML when N = 200, and to some extent when N = 300, but that direct ML outperforms pairwise ML at larger sample sizes.5 This finding replicates that of Enders (2001), who also found that pairwise ML seemed to do better than direct ML with highly nonnormal data in small sample sizes.”
• Robust Two-Stage Approach Outperforms Robust Full Information Maximum Likelihood With Incomplete Nonnormal Data (2014, methodology): “The goal of this study was to compare the relatively new robust TS approach for estimating and fitting structural equation models with incomplete nonnormal data to the best currently available analytic method for such data, robust FIML (Enders, 2001; Gold et al., 2003; Savalei & Bentler, 2005).”