A Comparison of Nonparametric Statistics and Bootstrap Methods for Testing Two Independent Populations with Unequal Variance

Article Sidebar

PDF
DOI: https://doi.org/10.28924/2291-8639-21-2023-36
Published: Apr 17, 2023
Cite as:
Wandee Wanishsakpong, Kantima Sodrung, Ampai Thongteeraparp, A Comparison of Nonparametric Statistics and Bootstrap Methods for Testing Two Independent Populations with Unequal Variance, Int. J. Anal. Appl., 21 (2023), 36.

Main Article Content

Wandee Wanishsakpong, Kantima Sodrung, Ampai Thongteeraparp

Abstract

The common parametric statistics used for testing two independent populations have often required the assumptions of normality and equal variances. Nonparametric tests have been used when assumptions of parametric tests cannot be achieved. However, some studies found nonparametric tests to be too conservative and less powerful than parametric tests. Bootstrap methods are also alternative tests when assumptions of parametric tests are violated, but they have small size limitations. Later, nonparametric tests when pooled with the bootstrap methods may overcome the powerful test and small sample sizes issue. Thus, the purpose of this study was to apply the bootstrap method together with nonparametric statistics and compare the efficiency of nonparametric tests and bootstraps methods when pooled with nonparametric tests for testing the mean difference between two independent populations with unequal variance. The Yuen Welch Test (YW), Brunner-Munzel Test (BM), Bootstrap Yuen Welch Test (BYW) and Bootstrap Brunner-Munzel Test (BBM) were studied via Monte Carlo simulation with non-normal population distributions. The results show that the probability of a type I error of all four test statistics could be controlled for all situations. The Brunner-Munzel test (BM) had the highest power and the best efficiency in the case of mean difference ratio increases. The Bootstrap Yuen Welch Test (BYW) had the highest power when the sample size was small.

Article Details

References

  1. J.V. Bradley, Robustness?, Br. J. Math. Stat. Psychol. 31 (1978), 144-152. https://doi.org/10.1111/j.2044-8317.1978.tb00581.x.
  2. E. Brunner, U. Munzel, The Nonparametric Behrens-Fisher Problem: Asymptotic Theory and a Small-Sample Approximation, Biom. J. 42 (2000), 17-25. https://doi.org/10.1002/(sici)1521-4036(200001)42:1%3C17::aid-bimj17%3E3.0.co;2-u.
  3. A.K. Dwivedi, I. Mallawaarachchi, L.A. Alvarado, Analysis of Small Sample Size Studies Using Nonparametric Bootstrap Test With Pooled Resampling Method, Stat. Med. 36 (2017), 2187-2205. https://doi.org/10.1002/sim.7263.
  4. B. Efron, R.J. Tibshirani, An Introduction to the Bootstrap, Chapman & Hall, New York, (1993).
  5. M.W. Fagerland, L. Sandvik, Performance of Five Two-Sample Location Tests for Skewed Distributions With Unequal Variances, Contemp. Clinic. Trials. 30 (2009), 490-496. https://doi.org/10.1016/j.cct.2009.06.007.
  6. W. Noppadon, B. Chinnapong, A Comparison Study of Nonparametric Test Statistics for Two Populations in Case of Heterogeneity of Variances, J. Sci. Technol. 18 (2010), 60-69.
  7. S. Ratiwat, Comparison of Nonparametric Statistical Test for Two Independent Difference Medians in Case of Symmetrical and Skewed Distribution, Master Thesis, Bangkok: King Mongkut’s Institute of Technology Ladkrabang Thailand, (2019).
  8. J. Reiczigel, I. Zakarias, L. Rozsa, A Bootstrap Test of Stochastic Equality of Two Populations, Amer. Stat. 59 (2005), 156-161. https://doi.org/10.1198/000313005x23526.
  9. R.R. Wilcox, Introduction to robust estimation and hypothesis testing (2nd ed.), Academic Press, San Diego, (2005).
  10. K.K. Yuen, The Two-Sample Trimmed T for Unequal Population Variances, Biometrika, 61 (1974), 165-170.
  11. D.W. Zimmerman, B.D. Zumbo, Rank Transformations and the Power of the Student t Test and Welch t’ Test for Non-Normal Populations With Unequal Variances, Canadian J. Exper. Psychol. 47 (1993), 523-539. https://doi.org/10.1037/h0078850.