Winsorization approach in testing the equality of independent groups

In testing the equality of independent groups, the ordinary approaches use classical tests such as t-test and analysis of variance (ANOVA) F test. However, these tests are known to have two common weaknesses namely nonnormality and heteroscedasticity. The Alexander-Govern test is acknowledged as a g...

Full description

Saved in:
Bibliographic Details
Main Authors: Faridzah, Jamaluddin, Suhaida, Abdullah, Yahaya Sharipah Soaad Syed
Format: Conference Proceeding
Language:English
Subjects:
Normality
Performance enhancement
Variance analysis
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In testing the equality of independent groups, the ordinary approaches use classical tests such as t-test and analysis of variance (ANOVA) F test. However, these tests are known to have two common weaknesses namely nonnormality and heteroscedasticity. The Alexander-Govern test is acknowledged as a good alternative to these tests in the presence of heteroscedasticity, but not in the case of considerable deviation from normality. Modifications were done to improve the performance of this test across various conditions, but the progress was still below satisfaction especially under heavy tailed distribution. For that reason, this study proposed a new approach of the Alexander-Govern test which employed the process of Winsorization. This process aims to eliminate the influence of nonnormality and unequal variances simultaneously. The results show that the proposed method performs better than the original method especially when the distribution is heavy tailed.
ISSN:0094-243X
1551-7616
DOI:10.1063/1.4887738