Comparison of type I error rates between T1 and Ft statistics for unequal population variance using variable trimming
Two robust procedures for testing the equality of central tendency measures, namely T1 and trimmed F (noted as Ft) statistics are proposed in this paper.The T1 and Ft statistics were modified using variable trimming with indeterminate percentage. The variable trimming percentages were based upon tri...
Saved in:
| Main Authors: | , , |
|---|---|
| 格式: | Article |
| 語言: | English |
| 出版: |
Universiti Putra Malaysia
2010
|
| 主題: | |
| 在線閱讀: | http://repo.uum.edu.my/19426/1/MJMS%204%202%20195-207%202010.pdf http://repo.uum.edu.my/19426/ http://einspem.upm.edu.my/journal/fullpaper/vol4no2/5.%20zahayu%20for%20web.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Two robust procedures for testing the equality of central tendency measures, namely T1 and trimmed F (noted as Ft) statistics are proposed in this paper.The T1 and Ft statistics were modified using variable trimming with indeterminate percentage. The variable trimming percentages were based upon trimming criteria using robust scale estimators, MADn and Tn.Altogether there are four procedures investigated: T1 with MADn, T1 with Tn, Ft with MADn, and Ft with Tn.Concentrating on just balanced design and unequal population variances, the four procedures were tested for their Type I error under different types of distributional shapes and total sample sizes.This study used 5000 simulated data sets to generate the Type I error. Since T1 distribution is unknown, bootstrap method was employed to test the hypothesis.The findings showed that T1 statistic works well under normal tail distribution, while Ft statistic is good for extremely skewed distribution. |
|---|