Comparing the "typical score" across independetn groups based on different criteria for trimming

Syed Yahaya, S. S. (avtor), Othman, A. R. (avtor), Keselman, H. J. (avtor)

Nonnormality and variance heterogeneity affect the validity of the traditionaltests for treatment group equality (e.g. ANOVA F-test and t-test), particularly when group sizes are unequal. Adopting trimmed means instead of the usual least squares estimator has been shown to be mostly affective in combating the deleterious effects of nonnormality. There are, however, practical concerns regarding trimmed means, such as the predetermined amount of symmetric trimming that is typically used. Wilcox and Keselman proposed theModified One- Step M-estimator (MOM) which empirically determines the amount of trimming. Othman et al. found that when this estimator is used with Schrader and Hettmanspergerćs H statistic, rates of Type I error were well controlled even though data were nonnormal in form. In this paper, we modifiedthe criterion for choosing the sample values for MOM by replacing the default scale estimator, MADn, with two robust scale estimators, Sn and Tn , suggested by Rousseeuw and Croux (1993). To study the robustness of the modified methods, conditions that are known to negatively affect rates of TypeI error were manipulated. As well, a bootstrap method was used to generatea better approximate sampling distribution since the null distributionof MOM-H is intractable. These modified methods resulted in betterType I error control especially when data were extremely skewed.

• Jezik: Angleški jezik
• Vrsta gradiva: Delo ni kategorizirano
• Tipologija: 1.01 - Izvirni znanstveni članek
• Organizacija: FDV - Fakulteta za družbene vede
• Leto izida: 2006
• Št. strani: Str. 49-62
• Številčenje: Vol. 3, no. 1
• PID: 20.500.12556/RUL-8441
• UDK: 303
• ISSN pri članku: 1854-0023
• COBISS.SI-ID: 25330525
• Datum objave v RUL: 11.07.2014

Gradivo je del revije

• Naslov: Metodološki zvezki
• Skrajšan naslov: Metodol. zv.
• Založnik: Fakulteta za družbene vede
• ISSN: 1854-0023
• COBISS.SI-ID: 215795712