Normal dağılım ve varyansların homojenliği ön şartlarının gerçekleşmediği durumlarda varyans analizi tekniğinin yerine kullanılabilecek bazı parametrik testlerin I. Tip hata ve testin gücü bakımından irdelenmesi
Özet
in (his study, when two basic assumptions of variance analysis, normality and homogeneity of variance, are not fullfilled. Type I error rales and power of test of some parametric tests (Marascuilo test, Welch test, Brown-Forsythc test, James second-ordet lest- Wilcox Hm lest and Alexander-Govern test) which can be used instead of ANOVA-F test were compared. The random samples were obtained from Normal, symetric (Beta (5,5), 1(5)) and various skew distributions (Beta (5,1), Beta (1,5), Weibull (i.5,1). Gamma (1,2), x*(3» by simulation. The comparisons were done by using these samples.At the end of 100 000 simulation trials, it was determined that these alternative tesis were belter than ANOVA-F test İn relation to Type I error rales and obtaining of adequate power of test, when especially variances were heterogenous. However, this circumtance was not valid for every experiment conditions. It was found tliat the actual Typo 1 error rate and power of lest for ANOVA-F test changed depending on population variances, sample size, observations balanced or not and the relation between population variance and sample size (direct pairing, inverse pairing) rather than disribulion shape of populations. The results also showed that in addition lu factors effecting ANOVA-F test, die distribution of shape effected these alternative tests.