Algorithm-based distribution of two-sample statistics

ABSTRACT

In this article, we develop an algorithm for working out the distribution of proposed test statistics. The computed distribution is compared with its existing asymptotic normal approximation using convergence rates and disparity measures. Simulation study is made to estimate the level of significance, and power to assess the performance of algorithm based procedure in comparison to asymptotic normal approximation. Both power comparisons and convergence rates suggest that the distribution computed through algorithm procedure be used for smaller sample sizes.

KEYWORDS:

• Disparity measure
• power comparison
• probability distribution
• recurrence relation

AMS SUBJECT CLASSIFICATION:

• 11K
• 49C
• 60E
• 62G

Acknowledgment

The authors gratefully acknowledge the editor and referees for their help in improving the article.