Usanee Janthasuwan.. Inferences for the parameters of skewed distributions. Doctoral Degree(Applied Statistics). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2024.
Inferences for the parameters of skewed distributions
Abstract:
This thesis presents the construction of confidence intervals for key parameters
of the delta-lognormal and delta-Birnbaum-Saunders distributions. These distributions are of particular interest as they combine right-skewed data with zero values that follow a binomial distribution, a key characteristic of both distributions. The parameters studied include the median, which serves as a robust measure of central tendency that is resistant to skewed data and outliers. Since the median is equivalent to the 50th percentile of a dataset, additional studies on percentiles were conducted to provide a broader statistical perspective. Additionally, the coefficient of variation is another important parameter used to measure relative dispersion and compare the distribution of datasets. Therefore, this thesis focuses on constructing confidence intervals for the median of the delta-lognormal distribution, including the analysis of differences and ratios of medians. The study extends to constructing simultaneous confidence intervals for the differences between medians and ratios of percentiles in the delta-lognormal
distribution. Furthermore, the thesis explores the delta-Birnbaum-Saunders distribution
by constructing confidence intervals for the coefficient of variation and extending the
study to the differences and ratios of coefficients of variation, as well as constructing
confidence intervals for the common coefficient of variation of several delta-Birnbaum-Saunders distributions. The performance of the constructed confidence intervals is evaluated based on coverage probability and average width, calculated through Monte Carlo simulations. Additionally, real data from rainfall and wind speed in Thailand are used to demonstrate the effectiveness of the proposed methods.