PERFORMANCE OF FDR TEST: A SIMULATION STUDY

Yıl 2013, Cilt: 1 Sayı: 2, 0 - 0, 01.10.2013

Aybüke Koca İbrahim Kılıç İsmet Doğan Sinan Saraçlı

Öz

Multiple comparison and multiple confidence tests are used to determine which group or groups are significantly different than others in analysis of variance. According to some assumptions, there are different kinds of multiple comparison and multiple confidence tests in the literature. In this study performance of the FDR (False Discovery Rate) test, which is one of the alternative test for the multiple comparison tests are examined. The significance value according to FDR is compared with the significance value for the t test. 3, 5 and 10 groups for the sample sizes 50, 100 and 200 are compared. MATLAB software is used to obtain the related data sets and to conclude the analysis. All of the results show thatFDR is nonsensitive to the number of compared groups. The increase of FDR values is constant for numbers of compared groups even if the sample size increases. Other results are given in related tables.

Anahtar Kelimeler

FDR, Multiple Comparison, Analysis of Variance