Search

Natural rainfall is necessary for agriculture in Thailand. Often, rainfall data contain zero and positive right-skewed observations. Within a given region, the mean rainfall can be used to evaluate how rainfall has changed over a period of time, and so, in this article, we propose interval estimates and an adjustment process based on the Bayesian approach to compute the rainfall amount mean. This includes highest posterior density intervals (HPDs) based on the beta (HPD-B), normal inverse chi-squared (HPD-NIC) and uniform (HPD-U) priors, which were compared with the existing methods. Coverage probability and relative average length were used to assess the performance of the methods by comparing their computation. A numerical evaluation showed that for a single mean and even chance of having zero observations, HPD-beta achieved the given target with small to moderate sample sizes, while HPD-U tended to perform very well with large sample size. To compare the difference between two means, HPD-U demonstrated excellent performance in almost all cases. Daily rainfall data from provinces in northern Thailand were used to confirm the efficacy of the new methods. [ABSTRACT FROM AUTHOR]

This paper proposes the novel approaches to construct confidence intervals for the common coefficient of variation (CV) of several normal populations using the concepts of the adjusted generalized confidence interval (adjusted GCI) approach and the computational approach. The coverage probability and average length of the proposed approaches were evaluated by a Monte Carlo simulation and compared with that of the existing approaches; the generalized confidence interval (GCI) approach and the adjusted method of variance estimates recovery (adjusted MOVER) approach. The results showed that the coverage probability of the adjusted GCI approach is above the nominal confidence level of 0.95 and better than the other approaches when the sample case is small ( k ≤ 6) for all sample sizes, except when the sample sizes are very small. However, the coverage probability of the computational approach provides much better confidence interval estimate than the other approaches when the sample case is large ( k ≥ 10). The proposed approaches and the existing approaches are illustrated using three medical science data. [ABSTRACT FROM AUTHOR]

Abstract Herein, we propose eight Bayesian methods (credible and highest posterior density intervals using the Jeffreys rule, independence Jeffreys, uniform, or normal-gamma priors) and one based on the fiducial quantity for constructing confidence intervals for the ratio of the means of zero-inflated gamma distributions. Simulation studies were conducted to evaluate the efficacies of the confidence intervals and compare their performances in terms of coverage probabilities and expected lengths, in which the Bayesian highest posterior density interval using the Jeffreys rule prior performed the best. We also applied the methods to rainfall datasets from northern Thailand to demonstrate their efficacies with real data. [ABSTRACT FROM AUTHOR]