Degree Type

Dissertation

Date of Award

1989

Degree Name

Doctor of Philosophy

Department

Statistics

First Advisor

Noel A. C. Cressie

Abstract

Several problems related to estimation, analysis and design with correlated observations are addressed. Under a one-dimensional covariance structure, asymptotic expressions of the bias of OLS residuals-based estimators for the covariance function and the variogram are presented. An estimator for the covariance function based on recursive residuals is introduced and compared with the classical OLS residuals-based estimator. A spatial approach to the analysis of experiments, where spatial dependence is assumed, is compared to the traditional one based on uncorrelated observations. The efficiency of incomplete block designs under second-order autoregressive error processes is studied and conditions for the universal optimality of balanced incomplete block designs are derived. Finally, results on the efficiency of first-order nearest-neighbor balanced designs are presented and its robustness against misspecification of a first-order autoregressive error process is discussed.

DOI

https://doi.org/10.31274/rtd-180813-8976

Publisher

Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/

Copyright Owner

Martin O. Grondona

Language

en

Proquest ID

AAI9003523

File Format

application/pdf

File Size

184 pages

Share

COinS