Type I Error and Power Rates of Ft Statistic with Trimmed Mean
Achieving nominal type I error rates and having high power values simultaneously will produce good test statistics. In order to identify a good test statistic which is able to satisfy both aforementioned criteria, a study is done on Ft statistic with trimming strategies using robust scale estimators...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Pushpa Publishing House
2012
|
| Subjects: | |
| Online Access: | https://repo.uum.edu.my/id/eprint/30957/1/FEJMS%2069%2001%202012%2037-50.pdf https://repo.uum.edu.my/id/eprint/30957/ http://pphmj.com/journals/fjms.htm |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Utara Malaysia |
| Language: | English |
| Summary: | Achieving nominal type I error rates and having high power values simultaneously will produce good test statistics. In order to identify a good test statistic which is able to satisfy both aforementioned criteria, a study is done on Ft statistic with trimming strategies using robust scale estimators, namely, MADn, Tn and LMSn. To test for the robustness of the procedures towards the violation of the assumptions, several variables are manipulated. The variables are types of distributions, heterogeneity of variances, sample sizes, nature of pairings of group sample sizes and group variances, and number of groups. This study is based on simulated data with each procedure simulated 5000 times. When testing for the hypothesis of the equality of central tendency measures, approximation method is used on Ft statistic. Type I error and power rates on J = 4 groups are then compared. Normal and skewed data from g- and h-distributions are considered in this study. Generally, all trimming strategies produce good type I error rates with high power values concurrently |
|---|