Relative power performance of t-test and bootstrap procedure for two-sample

• Main Authors: Ahad, Nor Aishah, Abdullah, Suhaida, Lai, Choo Heng, Mohd. Ali, Nazihah
• Format: Article
• Language: English
• Published: Universiti Putra Malaysia Press 2012
• Online Access: http://psasir.upm.edu.my/id/eprint/40432/1/4.%20Relative%20power%20performance%20of%20t-test%20and%20bootstrap%20procedure%20for%20two-sample.pdf; http://psasir.upm.edu.my/id/eprint/40432/; http://www.pertanika.upm.edu.my/Pertanika%20PAPERS/JST%20Vol.%2020%20%281%29%20Jan.%202012/09%20Pg%2043-52.pdf

The classical procedures of comparing two groups, such as t-test are, usually restricted with the assumptions of normality and equal variances. When these assumptions are violated, the rates of the Type I errors of the independent samples t-test are affected, particularly when the sample sizes are small. In this situation, the bootstrap procedure has an advantage over the parametric t-test. In this study, the performances of the bootstrap procedure and the independent sample t-test were investigated. The investigation focused on the power of both the test procedures to compare the two groups under different design specifications for normal and chi-square distributions. The results showed that the bootstrap procedure has a slight edge over the conventional t-test in term of the rate of achieving the benchmark level for both the distributions. In fact, the bootstrap procedure consistently outperformed the conventional t-test across all the combinations of the test conditions.