Abstract
This dissertation derives asymptotic properties of some estimators associated with spatial autoregressive processes. Explicit expressions for the asymptotic bias of Yule-Walker and Tjostheim's least squares estimators are obtained. The bias in the Yule-Walker estimators is shown to disappear if we use the so-called unbiased sample autocovariance function. The usual least squares estimators are also asymptotically normal and unbiased. The estimators for the frequency and the period at which the spectral density of spatial autoregressive processes has peaks are shown to be consistent and asymptotically normal. This dissertation also presents a statistical solution to problems in the field of meteorology, especially problems occurring in rain rate estimation. First, the statistical properties of the beam filling error are investigated based on North's approximation formula. The validation of North's approximation formula is demonstrated via simulation study. Second, the estimator of the optimal threshold level is examined and is shown to be consistent and asymptotically normal.
Ha, Eunho (1992). Analysis of spatial autoregressive processes and rain rate estimation. Texas A&M University. Texas A&M University. Libraries. Available electronically from
https : / /hdl .handle .net /1969 .1 /DISSERTATIONS -1354091.