Bivariate Correlation with Spatial Data

The paper examines the use of the Pearson product moment correlation coefficient and the Spearman rank correlation coefficient on map data that are spatially correlated. Research has shown that in this case standard estimators underestimate the true sampling variance of the Pearson coefficient. The Clifford and Richardson “effective sample size” solution to this problem is described, methods for implementing the procedure are considered, and properties of the technique examined. The method is also used to adjust the Spearman test. Two applications are given and the method of Clifford and Richardson contrasted with the method of data “prewhitening.”