Stephen G. West

A Comparison of Methods to Test Mediation and Other Intervening Variable Effects

A Monte Carlo study compared 14 methods to test the statistical significance of the intervening variable effect. An intervening variable (mediator) transmits the effect of an independent variable to a dependent variable. The commonly used R. M. Baron and D. A. Kenny (1986) approach has low statistical power. Two methods based on the distribution of the product and 2 difference-in-coefficients methods have the most accurate Type I error rates and greatest statistical power except in 1 important case in which Type I error rates are too high. The best balance of Type I error and statistical power across all cases is the test of the joint significance of the two effects comprising the intervening variable effect.

The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis.

Monte Carlo computer simulations were used to investigate the performance of three X 2 test statistics in confirmatory factor analysis (CFA). Normal theory maximum likelihood )~2 (ML), Brownes asymptotic distribution free X 2 (ADF), and the Satorra-Bentler rescaled X 2 (SB) were examined under varying conditions of sample size, model specification, and multivariate distribution. For properly specified models, ML and SB showed no evidence of bias under normal distributions across all sample sizes, whereas ADF was biased at all but the largest sample sizes. ML was increasingly overestimated with increasing nonnormality, but both SB (at all sample sizes) and ADF (only at large sample sizes) showed no evidence of bias. For misspecified models, ML was again inflated with increasing nonnormality, but both SB and ADF were underestimated with increasing nonnormality. It appears that the power of the SB and ADF test statistics to detect a model misspecification is attenuated given nonnormally distributed data.

Teacher's Corner: Testing Measurement Invariance of Second-Order Factor Models.

We illustrate testing measurement invariance in a second-order factor model using a quality of life dataset (n = 924). Measurement invariance was tested across 2 groups at a set of hierarchically structured levels: (a) configural invariance, (b) first-order factor loadings, (c) second-order factor loadings, (d) intercepts of measured variables, (e) intercepts of first-order factors, (f) disturbances of first-order factors, and (g) residual variances of observed variables. Given that measurement invariance at the factor loading and intercept levels was achieved, the latent factor mean difference on the higher order factor between the groups was also estimated. The analyses were performed on the mean and covariance structures within the framework of the confirmatory factor analysis using the LISREL 8.51 program. Implications of second-order factor models and measurement invariance in psychological research were discussed.

A Comparison of Bifactor and Second-Order Models of Quality of Life.

Bifactor and second-order factor models are two alternative approaches for representing general constructs comprised of several highly related domains. Bifactor and second-order models were compared using a quality of life data set (N = 403). The bifactor model identified three, rather than the hypothesized four, domain specific factors beyond the general factor. The bifactor model fit the data significantly better than the second-order model. The bifactor model allowed for easier interpretation of the relationship between the domain specific factors and external variables, over and above the general factor. Contrary to the literature, sufficient power existed to distinguish between bifactor and corresponding second-order models in one actual and one simulated example, given reasonable sample sizes. Advantages of bifactor models over second-order models are discussed.

The Problem of Units and the Circumstance for POMP.

Many areas of the behavioral sciences have few measures that are accepted as the standard for the operationalization of a construct. One consequence is that there is hardly ever an articulated and understood framework for the units of the measures that are employed. Without meaningful measurement units, theoretical formulations are limited to statements of the direction of an effect or association, or to effects expressed in standardized units. Thus the long term scientific goal of generation of laws expressing the relationships among variables in scale units is greatly hindered. This article reviews alternative methods of scoring a scale. Two recent journal volumes are surveyed with regard to current scoring practices. Alternative methods of scoring are evaluated against seven articulated criteria representing the information conveyed by each in an illustrative example. Converting scores to the percent of maximum possible score (POMP) is shown to provide useful additional information in many cases.

The science of prevention : methodological advances from alcohol and substance abuse research

The goal of The Science of Prevention: Methodological Advances From Alcohol and Substance Abuse Research is to promote critical thinking among new and established investigators about how to design research and analyze research findings. Although the substantive focus of many chapters is on applications to the prevention of alcohol and substance abuse, nearly all of the methodological principles and statistical models are general and have potential application to the full range of areas in which prevention research takes place. The contributors to this book share their knowledge from an informed, applied perspective. Most are active researchers in the field of substance abuse prevention who are also methodological experts. They have a firsthand knowledge not only of the methodological, statistical, and measurement issues but also of the substantive issues of their field. [publisher description]

Alternatives to the randomized controlled trial

Public health researchers are addressing new research questions (e.g., effects of environmental tobacco smoke, Hurricane Katrina) for which the randomized controlled trial (RCT) may not be a feasible option. Drawing on the potential outcomes framework (Rubin Causal Model) and Campbellian perspectives, we consider alternative research designs that permit relatively strong causal inferences. In randomized encouragement designs, participants are randomly invited to participate in one of the treatment conditions, but are allowed to decide whether to receive treatment. In quantitative assignment designs, treatment is assigned on the basis of a quantitative measure (e.g., need, merit, risk). In observational studies, treatment assignment is unknown and presumed to be nonrandom. Major threats to the validity of each design and statistical strategies for mitigating those threats are presented.

Putting the individual back into individual growth curves

Scaling of time (age) in latent growth curve (LGC) models has important implications for studies of development. When participants begin a study at different ages, sample means and covariance-based structural equation modeling (SEM) approaches produce biased estimates of the variance of the intercept and the covariance between the Intercept and Slope factors. However, individual data vector-based SEM approaches produce proper estimates of these parameters that are identical to those produced by multilevel modeling (MLM). Scaling of the time variable also raises issues regarding the interpretation of within- and between-persons effects of time that parallel those associated with centering of predictor variables in MLM. A numerical example is used to illustrate these issues, and an Mx script for fitting individual data vector-based LGC models is provided.

What the Need for Closure Scale Measures and What It Does Not: Toward Differentiating Among Related Epistemic Motives

The Need for Closure Scale (NFCS; D. M. Webster & A. W. Kruglanski, 1994) was introduced to assess the extent to which a person, faced with a decision o r judgment, desires any answer, as compared with confusion and ambiguity. The NFCS was presented as being unidimensional and as having adequate discriminant validity. Our data contradict these conceptual and psychometric claims. As a unidimensional scale, the NFCS is redundant with the Personal Need for Structure Scale (PNS; M. M. Thompson, M. E. Naccarato, & K. E. Parker, 1989). When the NFCS is used more appropriately as a multidimensional instrument, 3 of its facets are redundant with the PNS Scale, and a 4th is redundant with the Personal Fear of Invalidity Scale (M. M. Thompson et al., 1989). It is suggested that the NFCS masks important distinctions between 2 independent epistemic motives: the preference for quick, decisive answers (nonspecific closure) and the need to create and maintain simple structures (one form o f specific closure).

Health Beliefs and Compliance With Mammography-Screening Recommendations in Asymptomatic Women

The utility of the health belief model (HBM) for predicting compliance with the American Cancer Society recommendations for mammography screening over and above demographics, knowledge, physician input, and objective risk for breast cancer was assessed. In all, 615 predominantly middle-class White women, age 35-92, were surveyed in 1987-1989. A multiple indicator measurement model of the HBM constructs of perceived susceptibility, severity, benefits, and barriers was verified with confirmatory factor analysis. Physician input alone accounted for 25% of the variance in compliance; HBM constructs alone, 16%. HBM constructs accounted for 7% additional variance in compliance beyond all other predictors and thus may be a fruitful focus for interventions to increase screening rates.