ABSTRACT
This chapter explores the potentials of parallel processing in spatial statistical analysis. It describes in detail the parallel procedures to implement major operations in spatial autocorrelation analysis and spatial autoregressive modeling. A preliminary analysis of performance is also included to demonstrate the range of speedup we are able to achieve with parallel computers. The chapter provides a brief introduction to Single Instruction Multiple Data and Multiple Instruction Multiple Data architectures and the corresponding software models. It outlines the computational characteristics of the statistics and the procedures for implementing them on parallel computers. The chapter presents comparisons of processing times on a sequential and a data parallel platform, to show the potentials of parallel processing in spatial statistical analysis. It demonstrates that the data parallel model can efficiently implement all the operations in spatial statistics defined in two primary areas: detecting spatial patterns with spatial autocorrelation coefficients and modeling spatial processes with spatial autoregressive specifications.
