Testing the equality means of several log-normal distributions

ABSTRACT

Researches propose various methods for comparing the means of two log-normal distributions. Some of these methods have been recently extended to test the equality means of several log-normal populations. Investigations show that none of the established methods is satisfactory. In this article, we provide three methods based on the computational approach test, which is a parametric bootstrap approach, for testing the means of several log-normal distributions. Further, we compare our methods with the existing methods through Monte Carlo simulation. The numerical results show that the Type I errors of these procedures are satisfactory regardless of the sample size, number of populations, and the true parameters. Finally, we explain the considered methods by real examples.

KEYWORDS:

• Computational approach test
• Generalized p-value
• Log-normal distribution
• Maximum likelihood estimation
• Type I error.

MATHEMATICS SUBJECT CLASSIFICATION:

• 62F03
• 62F40