Bayesian sample size determination for coefficient of variation of normal distribution

Abstract

Sample size determination is an active area of research in statistics. Generally, Bayesian methods provide relatively smaller sample sizes than the classical techniques, particularly average length criterion is more conventional and gives relatively small sample sizes under the given constraints. The objective of this study is to utilize major Bayesian sample size determination techniques for the coefficient of variation of normal distribution and assess their performance by comparing the results with the freqentist approach. To this end, we noticed that the average coverage criterion is the one that provides relatively smaller sample sizes than the worst outcome criterion. By comparing with the existing frequentist studies, we show that a smaller sample size is required in Bayesian methods to achieve the same efficiency.

Keywords:

• Bayesian sample size
• coefficient of variation
• average coverage
• average length criterion
• worst outcome criterion

Acknowledgments

The authors are grateful to associate editor and two anonymous reviews for their constructive comments to improve the quality and presentation of the work.

Disclosure statement

No potential conflict of interest was reported by the author(s).