A study on the unequal probability sampling scheme and its application
LE3 .A278 2020
Master of Science
Mathematics & Statistics
Sampling is an indispensable part of statistics. It helps us to study the characteristics of the whole target population without investigating every individual of the population, which largely economizes time and cost and enhances efficiency. Various sampling strategies have raised in the past decades. The development and a brief explanation of the types of sampling schemes are provided in Chapter 1. Our study targets unequal probability sampling without replacement with fixed sample size. Five popular sampling schemes, including Poisson sampling, conditional Poisson sampling, Tille's sampling, Sampford's sampling and systematic sampling are introduced in Chapter 4, as well as a new sequential sampling scheme (Lu's method). In Chapter 3, comparisons between conditional Poisson sampling, Lu's method, Tille's Method and Sampford's sampling with different data sets are made in terms of different variance estimators and entropy. It is shown that Lu's method is a reliable and competitive sampling scheme. Chapter 4 applies Lu's method, conditional Poisson sampling, Poisson sampling and systematic sampling to a real-world problem and compares the performance of those methods. A comprehensive summary and a conclusion are presented in Chapter 5.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.