Survey and longitudinal studies in the social and behavioral sciences generally contain missing data. Mean and covariance structure models play an important role in analyzing such data. Two promising methods for dealing with missing data are a direct maximum-likelihood and a two-stage approach based on the unstructured mean and covariance estimates obtained by the EM-algorithm. Typical assumptions under these two methods are ignorable nonresponse and normality of data. However, data sets in social and behavioral sciences are seldom normal, and experience with these procedures indicates that normal theory based methods for nonnormal data very often lead to incorrect model evaluations. By dropping the normal distribution assumption, we develop more accurate procedures for model inference. Based on the theory of generalized estimating equations, a way to obtain consistent standard errors of the two-stage estimates is given. The asymptotic efficiencies of different estimators are compared under various assump...

Three Likelihood-Based Methods For Mean and Covariance Structure Analysis With Nonnormal Missing Data

This article reviews proposed goodness-of-fit indices for structural equation models and the Monte Carlo studies that have empirically assessed their distributional properties. The cumulative contributions of the studies are summarized, and the variables under which the indices are studied are noted. A primary finding is that many of the indices used until the late 1980s, including Joreskog and Sorbom's (1981) GFI and Bentler and Bonett's (1980) NFI, indicated better fit when sample size increased. More recently developed indices based on the chi-square noncentrality parameter are discussed and the relevant Monte Carlo studies reviewed. Although a more complete understanding of their properties and suitability requires further research, the recommended fit indices are the McDonald (1989) noncentrality index, the Bentler (1990)-McDonald and Marsh (1990) RNI (or the bounded counterpart CFI), and Bollen's (1989) DELTA2.

Monte Carlo Evaluations of Goodness of Fit Indices for Structural Equation Models

This simulation study demonstrates 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. Using a fully crossed design, data were generated for 11 conditions of peakedness, 3 conditions of misspecification, and 5 different sample sizes. Three estimation methods (maximum likelihood [ML], generalized least squares [GLS], and weighted least squares [WLS]) were compared in terms of overall fit and the discrepancy between estimated parameter values and the true parameter values used to generate the data. Consistent with earlier findings, the results show that ML compared to GLS under conditions of misspecification provides more realistic indexes of overall fit and less biased parameter values for paths that overlap with the true model. However, despite recommendations found in the literature that WLS should be used when data are not normally distribute...

The Performance of ML, GLS, and WLS Estimation in Structural Equation Modeling Under Conditions of Misspecification and Nonnormality

The concept of moderate consumption of ethanol (beverage alcohol) has evolved over time from considering this level of intake to be nonintoxicating and noninjurious, to encompassing levels defined as "statistically" normal in particular populations, and the public health-driven concepts that define moderate drinking as the level corresponding to the lowest overall rate of morbidity or mortality in a population. The various approaches to defining moderate consumption of ethanol provide for a range of intakes that can result in blood ethanol concentrations ranging from 5 to 6 mg/dl, to levels of over 90 mg/dl (i.e., approximately 20 mM). This review summarizes available information regarding the effects of moderate consumption of ethanol on the adult and the developing nervous systems. The metabolism of ethanol in the human is reviewed to allow for proper appreciation of the important variables that interact to influence the level of exposure of the brain to ethanol once ethanol is orally consumed. At the neurochemical level, the moderate consumption of ethanol selectively affects the function of GABA, glutamatergic, serotonergic, dopaminergic, cholinergic, and opioid neuronal systems. Ethanol can affect these systems directly, and/or the interactions between and among these systems become important in the expression of ethanol's actions. The behavioral consequences of ethanol's actions on brain neurochemistry, and the neurochemical effects themselves, are very much dose- and time-related, and the collage of ethanol's actions can change significantly even on the rising and falling phases of the blood ethanol curve. The behavioral effects of moderate ethanol intake can encompass events that the human or other animal can perceive as reinforcing through either positive (e.g., pleasurable, activating) or negative (e.g., anxiolysis, stress reduction) reinforcement mechanisms. Genetic factors and gender play an important role in the metabolism and behavioral actions of ethanol, and doses of ethanol producing pleasurable feelings, activation, and reduction of anxiety in some humans/animals can have aversive, sedative, or no effect in others. Research on the cognitive effects of acute and chronic moderate intake of ethanol is reviewed, and although a number of studies have noted a measurable diminution in neuropsychologic parameters in habitual consumers of moderate amounts of ethanol, others have not found such changes. Recent studies have also noted some positive effects of moderate ethanol consumption on cognitive performance in the aging human. The moderate consumption of ethanol by pregnant women can have significant consequences on the developing nervous system of the fetus. Consumption of ethanol during pregnancy at levels considered to be in the moderate range can generate fetal alcohol effects (behavioral, cognitive anomalies) in the offspring. A number of factors--including gestational period, the periodicity of the mother's drinking, genetic factors, etc.--play important roles in determining the effect of ethanol on the developing central nervous system. A series of recommendations for future research endeavors, at all levels, is included with this review as part of the assessment of the effects of moderate ethanol consumption on the central nervous system.

Effects of Moderate Alcohol Consumption on the Central Nervous System

Is there a menopausal syndrome? Menopausal status and symptoms across racial/ethnic groups.

Using the theory of pseudo maximum likelihood estimation the asymptotic covariance matrix of maximum likelihood estimates for mean and covariance structure models is given for the case where the variables are not multivariate normal This asymptotic covariance matrix is consistently estimated without the computation of the empirical fourth order moment matrix Using quasi-maximum likelihood theory a Hausman misspecification test is developed This test is sensitive to misspecification caused by errors that are correlated with the independent variables This misspecification cannot be detected by the test statistics currently used in covariance structure analysis

Pseudo maximum likelihood estimation and a test for misspecification in mean and covariance structure models

A representative sample of 1,367 employed men and women in Detroit responded to questions about their drinking practices and then completed a cognitive test which measures abstraction abilities. Abstraction, tested while respondents were sober, decreased significantly as reported quantity of alcohol usually consumed per drinking occasion increased. (Am J Public Health 1983; 73:521-526.)

Alcohol Use and Cognitive Loss among Employed Men and Women

In two samples of employed men (ns of 102 and 496), the amount of alcohol typically consumed per drinking occasion was significantly associated with decreased sober abstraction performance. Age was also significantly related to reduced abstraction scores. An increase of one drink in the quantity of alcohol typically consumed was associated with an average reduction in abstraction performance equivalent to 3.7 more yr of age in one sample and 2.4 more yr of age in the other sample.

Cognitive patterns resembling premature aging in male social drinkers.

This issue of Sociological Methods & Research contains four excellent illustrations of current methods for analyzing dichotomous data: dichotomous factor analysis (Muthen, 1989), latent trait analy...

Latent Variable Models of Dichotomous Data The State of the Method

The EM (Estimation-Maximization) algorithm is exploited to provide maximum likelihood estimates of the parameters of multiple indicator/factor analysis models. This method reduces considerably the storage and computational burden of such estimation. A computer program in BASIC language that performs the computations is listed in an appendix. The specification of correlated errors is also provided for in this application of the method.

Ronald Schoenberg

Papers

Pseudo maximum likelihood estimation and a test for misspecification in mean and covariance structure models

Alcohol Use and Cognitive Loss among Employed Men and Women

Cognitive patterns resembling premature aging in male social drinkers.

Latent Variable Models of Dichotomous Data The State of the Method

Application of the EM Method A Study of Maximum Likelihood Estimation of Multiple Indicator and Factor Analysis Models