A comparison of type I error and power of Bartlett’s test, Levene’s test and Cochran’s test under violation of assumptions
نویسندگان
چکیده
Vorapongsathorn, T., Taejaroenkul, S. and Viwatwongkasem, C. A comparison of type I error and power of Bartlett’s test, Levene’s test and Cochran’s test under violation of assumptions Songklanakarin J. Sci. Technol., 2004, 26(4) : 537-547 This study compared the probability of Type I error and the power of three statistical tests (Bartlett, Levene and Cochran) by varying the sampling distribution, variances and sample sizes. Monte Carlo methods were used to generate responses based on sample sizes and distributions 1,000 times. The sample sizes were both equal and unequal: 15, 30 and 45. The data distributions were Normal, Gamma and Chi-square. It was found that Bartlett’s test was sensitive to the normality assumption whereas Cochran’s test and Levene’s test were robust when the normal assumption was violated. Moreover, Levene’s test was quite good for both equal and small sample sizes. In the case of power, Bartlett’s test had the highest power in all cases. When one variance was large, Cochran’s test was the best test. 1Ph.D.(Research Design & Statistics), Assoc. Prof., 3M.Sc.(Biostatistics), Assoc. Prof., Department of Biostatistics, Faculty of Public Health, Mahidol University, Bangkok 10400 Thailand 2M.Sc.(Biostatistics), Statistician, Research Institute for Health Science, Chiang Mai University, Chiang Mai, 50200 Thailand. Corresponding e-mail: [email protected] Received, 8 January 2004 Accepted, 15 March 2004 ORIGINAL ARTICLE Songklanakarin J. Sci. Technol. Vol. 26 No. 4 Jul.-Aug. 2004 538 Comparison of type I error under violation of assumptions Vorapongsathorn, T., et al. The recommendations from this study are that: Bartlett’s test is the best test for homogeneity of variances since it is not affected by sample size. When data are non-normally distributed, Levene’s test is a good choice for small equal sample sizes. Cochran’s test is best when sample size is large and unequal and one variance is larger.
منابع مشابه
Modified signed log-likelihood test for the coefficient of variation of an inverse Gaussian population
In this paper, we consider the problem of two sided hypothesis testing for the parameter of coefficient of variation of an inverse Gaussian population. An approach used here is the modified signed log-likelihood ratio (MSLR) method which is the modification of traditional signed log-likelihood ratio test. Previous works show that this proposed method has third-order accuracy whereas the traditi...
متن کاملComparing two testing procedures in unbalanced two-way ANOVA models under heteroscedasticity: Approximate degree of freedom and parametric bootstrap approach
The classic F-test is usually used for testing the effects of factors in homoscedastic two-way ANOVA models. However, the assumption of equal cell variances is usually violated in practice. In recent years, several test procedures have been proposed for testing the effects of factors. In this paper, the two methods that are approximate degree of freedom (ADF) and parametric bootstr...
متن کاملEvaluation of global testing procedures for item fit to the Rasch model.
Two types of global testing procedures for item fit to the Rasch model were evaluated using simulation studies. The first type incorporates three tests based on first-order statistics: van den Wollenberg's Q(1) test, Glas's R(1) test, and Andersen's LR test. The second type incorporates three tests based on second-order statistics: van den Wollenberg's Q(2) test, Glas's R(2) test, and a non-par...
متن کاملStatistical comparison of univariate tests of homogeneity of variances
This paper compares empirical type I error and power of different tests that have been proposed to assess the homogeneity of within-group variances, prior to anova. The tests of homogeneity of variance (THV) compared in this study are: Bartlett's test, the Scheffé-Box log-anova test, Cochran’s C test and Box’s M test, in their parametric and permutational forms. The main questions addressed in ...
متن کاملThe effects of the violation of local independence assumption on the person measures under the Rasch model
Local independence of test items is an assumption in all Item Response Theory (IRT) models. That is, the items in a test should not be related to each other. Sharing a common passage, which is prevalent in reading comprehension tests, cloze tests and C-Tests, can be a potential source of local item dependence (LID). It is argued in the literature that LID results in biased parameter estimation ...
متن کامل