2007
DOI: 10.1007/s11135-007-9133-z
|View full text |Cite
|
Sign up to set email alerts
|

Evaluating estimation methods for ordinal data in structural equation modeling

Abstract: Estimation, Ordinal data, Model misspecification, Small sample structural equation modeling,

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
2

Citation Types

5
130
2
12

Year Published

2008
2008
2019
2019

Publication Types

Select...
6
1

Relationship

0
7

Authors

Journals

citations
Cited by 148 publications
(149 citation statements)
references
References 18 publications
(26 reference statements)
5
130
2
12
Order By: Relevance
“…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: 89%
See 2 more Smart Citations
“…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: 89%
“…In contrast, Yang-Wallentin, Jöreskog, and Luo (2010) gave empirical evidence that the parameter estimates (both factor loadings and interfactor correlations) obtained by WLS were substantially biased, whereas those obtained by WLSMV and ML were essentially unbiased, regardless of the number of categories (two, five, or seven) and the shape of the observed distributions (symmetrical vs. asymmetrical). In addition, Lei (2009) found that the relative bias in parameter estimates was generally negligible for both ML and WLSMVacross different levels of asymmetric observed distributions (symmetric, mildly skewed, and moderately skewed). Oranje (2003) also concluded that both ML and WLSMV produced equally good parameter estimates across the numbers of categories (two, three, and five).…”
Section: Previous Simulation Studiesmentioning
confidence: 94%
See 1 more Smart Citation
“…Respecto al aná-lisis cuantitativo, se aplicó un análisis factorial confirmatorio desde el modelamiento de ecuaciones estructurales (Structural Equation Modeling, SEM;Bentler y Dugeon, 1996;Jöreskog, 1969), mediante el cual se verificó la fuente de varianza latente de los ítems del BSSS. El método utilizado fue el de máxima verisimilitud con el escalamiento de Satorra y Bentler (1994;SB-c 2 ), pues es un procedimiento efectivo cuando ocurren distribuciones no normales de los ítems (Boomsma, 2000;Lei y Wu, 2012;Tong y Bentler, 2013), y permite aproximar mejor la prueba de bondad de ajuste a la distribución c 2 de (Bentler y Dugeon, 1996). Los análisis estructurales se basaron en la matriz de covarianzas, considerando que el número de alternativas de respuesta de los ítems (cinco), es una característica suficiente para aproximarse a variables continuas, sin que produzca sesgos sustanciales en los parámetros estimados aun usando el método maximun likehood (Beauducel y Herzberg, 2006;Dolan, 1994;Rhemtulla, Brosseau-Liard y Savalei, 2012).…”
Section: Procedimientounclassified
“…Ya que esta especificación inicial puede requerir ser relajada durante el análi-sis en un marco a posteriori (Boomsma, 2000), se tomaron dos criterios para ello: uno de tipo estadístico mediante el estudio de los índices de Lagrange (Sörbom, 1989), llamados también índices de modificación; y otro de tipo racional, el mismo que tiene una base conceptual y teórica, y que se considera relativamente más importante (Boomsma, 2000;Lei y Wu, 2012) (McDonald, 1989). Este conjunto de índices de ajuste es recomendado para ayudarse en la toma de decisiones sobre los modelos evaluados (Jackson, Gillaspay y Purc-Stephenson, 2009).…”
Section: Procedimientounclassified