Provides a comprehensive, nontechnical overview of the three maximum likelihood algorithms available for use with missing data and discusses multiple imputation, frequently used in conjunction with the EM algorithm. (SLD)

Developed an extension of the general latent variable growth curve modeling framework to four levels of the hierarchy. The extension merged two common analytical approaches: full information maximum likelihood (ML) latent growth modeling and limited information multilevel latent growth modeling using an ML estimator. Results for data from 250…

Demonstrated, through simulation, that stationary autoregressive moving average (ARMA) models may be fitted readily when T>N, using normal theory raw maximum likelihood structural equation modeling. Also provides some illustrations based on real data. (SLD)

A didactic discussion of covariance structure modeling in longitudinal studies with missing data is presented. Use of the full-information maximum likelihood method is considered for model fitting, parameter estimation, and hypothesis testing purposes, particularly when interested in patterns of temporal change as well as its covariates and…

An Expectation-Maximization (EM) algorithm in a maximum likelihood framework is developed to estimate finite mixtures of multivariate regression and simultaneous equation models with multiple endogenous variables. A dataset with cross-sectional observations for a diverse sample of businesses illustrates the semiparametric approach. (SLD)

Reexamined the nature of structural equation modeling (SEM) estimates of autoregressive moving average (ARMA) models, replicated the simulation experiments of P. Molenaar, and examined the behavior of the log-likelihood ratio test. Simulation studies indicate that estimates of ARMA parameters observed with SEM software are identical to those…

Used Monte Carlo simulation to examine the performance of four missing data methods in structural equation models: (1)full information maximum likelihood (FIML); (2) listwise deletion; (3) pairwise deletion; and (4) similar response pattern imputation. Results show that FIML estimation is superior across all conditions of the design. (SLD)

Studied the maximum likelihood estimation of unknown parameters in a general LISREL-type model with mixed polytomous and continuous data through Monte Carlo simulation. Proposes a model selection procedure for obtaining good models for the underlying substantive theory and discusses the effectiveness of the proposed model. (SLD)

Relative fit indices using the null model as the reference point in computation may differ across estimation methods, as this article illustrates by comparing maximum likelihood, ordinary least squares, and generalized least squares estimation in structural equation modeling. The illustration uses a covariance matrix for six observed variables…

Investigated the assumption that determining an adequate sample size in structural equation modeling can be aided by considering the number of parameters to be estimated. Findings from maximum likelihood confirmatory factor analysis support previous research on the effect of sample size, measured variable reliability, and the number of measured…

Two methods, direct maximum likelihood (ML) and the expectation maximization (EM) algorithm, can be used to obtain ML parameter estimates for structural equation models with missing data (MD). Although the 2 methods frequently produce identical parameter estimates, it may be easier to satisfy missing at random assumptions using EM. However, no…

Presents an application of latent growth curve methodology to the analysis of longitudinal developmental change in alcohol consumption of 586 young adults, illustrating three approaches to the analysis of missing data: (1) multiple-sample structural equation modeling procedures; (2) raw maximum likelihood analyses; and (3) multiple modeling and…

Used simulation to demonstrate how the choice of estimation method affects indexes of fit and parameter bias for different sample sizes when nested models vary in terms of specification error and the data demonstrate different levels of kurtosis. Discusses results for maximum likelihood (ML), generalized least squares (GLS), and weighted least…

Actual kurtotic and skewed data and varied sample sizes and estimation methods demonstrated that normal theory maximum likelihood and generalized least square estimators were fairly consistent and almost identical. Standard errors tended to underestimate the estimator's true variation but the problem was not serious for large samples. (SLD)

Used simulation to compare structural equation modeling methods for estimating and testing hypotheses about an interaction between continuous variables. Findings indicate that the two-stage least squares procedure exhibited more bias and lower power than the other methods. The Jaccard-Wan procedure (J. Jaccard and C. Wan, 1995) and maximum…