Date of Award
Doctor of Philosophy
Jae Kwang Kim
This dissertation consists of three parts. In the first part, we propose new bootstrap methods for three commonly used sampling designs, including the Poisson sampling, simple random sampling, and probability-proportional-to-size sampling. We show that the proposed bootstrap methods are second-order accurate and easy to be implemented in practice. Two simulation studies are conducted to compare the proposed bootstrap methods with the Wald method, and the proposed bootstrap methods outperform the Wald method in terms of coverage rate. It is well-known that a spatially balanced sample, which spread over the study domain well, can improve the estimation efficiency under dependent settings. In the second part, we propose to use a block bootstrap method to estimate the variance and make inference based on a sample generated by a one-per-stratum sampling design. We show the validity of the block bootstrap method and compare it with another commonly used sampling design theoretically. Simulation study shows that the block bootstrap
method can provide valid variance estimator and inference for the one-per-stratum sampling design. Although there are many researches about spatially balanced sampling design, there are few discussing the spatio-temporal balanced sampling design. In the third part, we propose a spatio-temporal balanced sampling design to generate annual samples, such that the sample for each year
is spatially balanced, and the one combining from consecutive years is also spatially balanced. We also propose design-based variance estimator for the estimates of annual status and annual change. The proposed sampling design is used in the National Resources Inventory rangeland on-site survey, and it shows that the proposed design performs better than the current design and estimators.
Wang, Zhonglei, "Topics in bootstrap methods for survey sampling and spatially balanced design" (2018). Graduate Theses and Dissertations. 17349.