Figure 3.

Quantitative effects of correction on spatial autocorrelation. Reimers-Weinstein spatial autocorrelation metric computed in data corrected by various methods (y-axis) and in original data (x-axis) for each array in each dataset, with the “selected” array from each dataset being emphasized by a solid bullet. The Reimers-Weinstein metric is a Pearson correlation coefficient computed between each intensity and the average intensity among its four neighbours on the array [10]. An unchanged metric lies on the dotted unit line, while a value below (above) this line indicates a decrease (increase) in spatial autocorrelation. Upton-Lloyd consistently results in the greatest decrease, followed by pyn, while CPP and LPE have no effect in all but one case.