Mardia's Multivariate Kurtosis with Missing Data

Yuan, Ke‐Hai; Lambert, Paul L.; Fouladi, Rachel T.

doi:10.1207/s15327906mbr3903_2

Cited by 62 publications

(37 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Of all the total participants (N = 485), 12% had a missing value, and among them there were 19 missing value patterns. As the Yuan, Lambert, and Fouladi (2004) coefficient of kurtosis standardized was high (75.45) the robust method was also applied to both the measurement model as well as the Yuan and Bentler (2000) structural model. As expected, in some 142 I. Ibabe univariate distributions a marked lack of normality was detected: mother-son/ daughter violence (asymmetry = 3.27, kurtosis = 12.15), son/daughter-parent physical violence (asymmetry = 2.66, kurtosis = 6.49) and son/daughter-parents financial violence (asymmetry = 2.7, kurtosis = 7.02).…”

Section: Discussionmentioning

confidence: 99%

Direct and indirect effects of family violence on child-to-parent violence / Efectos directos e indirectos de la violencia familiar sobre la violencia filio-parental

Ibabe

2014

Estudios de Psicología

View full text Add to dashboard Cite

show abstract

Section: Discussionmentioning

confidence: 99%

Direct and indirect effects of family violence on child-to-parent violence / Efectos directos e indirectos de la violencia familiar sobre la violencia filio-parental

Ibabe

2014

Estudios de Psicología

View full text Add to dashboard Cite

show abstract

“…The analysis was carried out with full information maximum likelihood (e.g., Arbuckle, 1996;Jamshidian & Bentler, 1998). While the maximum likelihood estimates were accepted, since the Yuan, Lambert, and Fouladi (2004) normalized coefficient of kurtosis exceeded 10.0 in all analyses, the associated model test statistics and standard errors were corrected for nonnormality using the Yuan and Bentler (2000) robust methodology. Although fit indices based on normal theory and robust statistics were very which they behaved violently against parents and teachers 1 on a 5-point Likert scale (See Appendix Scale).…”

Section: Discussionmentioning

confidence: 99%

Protective Factors for Adolescent Violence against Authority

2013

View full text Add to dashboard Cite

Both the family and school environments influence adolescents' violence, but there is little research focusing simultaneously on the two contexts. This study analyzed the role of positive family and classroom environments as protective factors for adolescents' violence against authority (parent abuse and teacher abuse) and the relations between antisocial behavior and child-to-parent violence or student-to-teacher violence. The sample comprised 687 Spanish students aged 12-16 years, who responded to the Family Environment Scale (FES) and the Classroom Environment Scale (CES). Structural Equation Modeling was used to test our model of violent behavior towards authority based on Catalano and Hawkins' Social Developmental Model (1996). Perceived family cohesion and organization showed an inverse association with parent abuse, suggesting that a positive family environment was a protective factor for the development of violence against parents. Family and classroom environments had direct effects on adolescents' violence against authority, and antisocial behavior showed a mediating effect in this relationship. The model accounted for 81% of the variance in violence against authority. As family environment was a better predictor of violence against authority than school environment, intervention efforts to reduce rates of adolescent violence should focus on helping parents to increase family cohesion and to manage conflictive relationships with their children.

show abstract

“…Em compensação, nenhum deles apresentou distribuição normal em termos univariados, mesmo após a sua transformação em escores z, o que poderia reduzir ou eliminar este problema (HAIR JR. et al, 2006). Consequentemente, a curtose multivariada foi de 9,06 no caso do ACSI e de 11,19 para o ECSI, valores superiores ao aceitável para se atestar a presença de normalidade multivariada, que é de 3 (HENZE, 1994;YUAN;LAMBERT;FOULADI, 2004). Violações a esse pressuposto não inviabilizam a utilização da análise multivariada, mas podem produzir resultados com algum grau de enviezamento (JÖRESKOG; SÖRBOM, 2000;GERBING, 1988 (TABACHNICK; FIDELL, 2001).…”

Section: Amostragem E Amostraunclassified