Confirmatory Factor Analysis of Ordinal Variables With Misspecified Models

Yang-Wallentin, Fan; Jöreskog, Karl G.; Luo, Hao

doi:10.1080/10705511.2010.489003

Cited by 246 publications

(236 citation statements)

References 16 publications

Supporting

Mentioning

208

Contrasting

Unclassified

Order By: Relevance

“…First, both estimators were not subject to the problems of improper solutions or nonconvergence with a small sample (N = 200) in a correlated two-factor model, consistent with previous simulation studies (Flora & Curran, 2004;Herzog, Boomsma, & Reinecke, 2007). Prior scholarship, however, has observed nonconvergence or improper solutions, in particular, when data were analyzed in quite small samples N = 100 or 150 (Rhemtulla, BrosseauLiard, & Savalei, 2012;Yang-Wallentin et al, 2010). Second, this study replicated previous results that factor loadings are typically underestimated by MLR but are essentially unbiased with WLSMV (Beauducel & Herzberg, 2006;DiStefano, 2002;Flora & Curran, 2004).…”

Section: Discussionmentioning

confidence: 52%

“…On the other hand, simulation studies have shown that standard errors in WLSMV were generally less biased than those obtained by meanadjusted ML, irrespective of the number of categories (Yang-Wallentin et al, 2010) and the level of asymmetric observed distributions (Lei, 2009). As for chi-square statistics, Beauducel and Herzberg (2006) revealed that the unadjusted chi-square statistics produced by ML were more likely to over-reject the proposed models than were the mean-and variance-adjusted chi-square statistics obtained by WLSMV.…”

Section: Previous Simulation Studiesmentioning

confidence: 91%

“…In terms of standard error estimates, ML has been found to produce much smaller standard errors of factor loadings than does mean-adjusted ML in LISREL, statistically equivalent to MLM in Mplus, (Yang-Wallentin et al, 2010) and WLSMV (Beauducel & Herzberg, 2006), indicating that uncorrected ML standard errors are generally underestimated. On the other hand, simulation studies have shown that standard errors in WLSMV were generally less biased than those obtained by meanadjusted ML, irrespective of the number of categories (Yang-Wallentin et al, 2010) and the level of asymmetric observed distributions (Lei, 2009).…”

Section: Previous Simulation Studiesmentioning

confidence: 99%

“…A correlated two-factor model was chosen as the representative of the common/simple CFA model specification that is oftentimes examined in simulation studies and is frequently encountered in practice. The literature in ordinal CFA is abundant for the joint performance of robust estimators on both factor loading and interfactor correlation estimates (e.g., Lei, 2009;Oranje, 2003;Yang-Wallentin et al, 2010) or the performance of non-robust estimators (e.g., DiStefano, 2002;). However, little is currently known about the performance of different estimators with statistical corrections when examining factor loadings and interfactor correlations separately, along with corrected standard errors and chisquare test statistics, as well.…”

Section: Present Studymentioning

confidence: 99%

See 3 more Smart Citations

Confirmatory factor analysis with ordinal data: Comparing robust maximum likelihood and diagonally weighted least squares

2015

Behav Res

1,962

1,162

View full text Add to dashboard Cite

In confirmatory factor analysis (CFA), the use of maximum likelihood (ML) assumes that the observed indicators follow a continuous and multivariate normal distribution, which is not appropriate for ordinal observed variables. Robust ML (MLR) has been introduced into CFA models when this normality assumption is slightly or moderately violated. Diagonally weighted least squares (WLSMV), on the other hand, is specifically designed for ordinal data. Although WLSMV makes no distributional assumptions about the observed variables, a normal latent distribution underlying each observed categorical variable is instead assumed. A Monte Carlo simulation was carried out to compare the effects of different configurations of latent response distributions, numbers of categories, and sample sizes on model parameter estimates, standard errors, and chi-square test statistics in a correlated two-factor model. The results showed that WLSM V was less biased and more accurate than MLR in estimating the factor loadings across nearly every condition. However, WLSMV yielded moderate overestimation of the interfactor correlations when the sample size was small or/and when the latent distributions were moderately nonnormal. With respect to standard error estimates of the factor loadings and the interfactor correlations, MLR outperformed WLSMV when the latent distributions were nonnormal with a small sample size of N = 200. Finally, the proposed model tended to be over-rejected by chi-square test statistics under both MLR and WLSMV in the condition of small sample size N = 200.

show abstract

Section: Discussionmentioning

confidence: 52%

Section: Previous Simulation Studiesmentioning

confidence: 91%

Section: Previous Simulation Studiesmentioning

confidence: 99%

Section: Present Studymentioning

confidence: 99%

See 2 more Smart Citations

Confirmatory factor analysis with ordinal data: Comparing robust maximum likelihood and diagonally weighted least squares

2015

Behav Res

1,962

1,162

View full text Add to dashboard Cite

show abstract

“…In WLS, the weight matrix is an estimate of the inverse of the asymptotic covariance matrix of polychoric correlations, while DWLS involves only the diagonal elements of that weight matrix. Recent studies confirm (Forero, Maydeu-Olivares, & Gallardo-Pujol, 2009;Yang-Wallentin, Jöreskog, & Luo, 2010) that the WLS estimator converges very slowly to its asymptotic properties and therefore does not perform well in small sample sizes. DWLS and ULS are preferable to WLS and they seem to perform similarly well in finite samples.…”

Section: Discussionmentioning

confidence: 98%

Analysis of Multivariate Longitudinal Data Subject to Nonrandom Dropout

Hafez

Moustaki

Kuha

2014

Structural Equation Modeling: A Multidisciplinary Journal

View full text Add to dashboard Cite

Longitudinal data are collected for studying changes across time. We consider multivariate longitudinal data where multiple observed variables, measured at each time point, are used as indicators for theoretical constructs (latent variables) of interest. A common problem in longitudinal studies is dropout, where subjects exit the study prematurely. Ignoring the dropout mechanism can lead to biased estimates, especially when the dropout is nonrandom. Our proposed approach uses latent variable models to capture the evolution of the latent phenomenon over time while also accounting for possibly nonrandom dropout. The dropout mechanism is modeled with a hazard function that depends on the latent variables and observed covariates. Different relationships among these variables and the dropout mechanism are studied via two model specifications. The proposed models are used to study people's perceptions on women's work using three questions from five waves from the British Household Panel Survey.

show abstract