Skip to main content
. 2014 Jan 31;9(1):e88119. doi: 10.1371/journal.pone.0088119

Figure 1. Schematic diagram of the simulation study.

Figure 1

The stimulation was conducted to determine the Window size of Moran’s I that best reflected the inflated topological correlation of the two simulated networks. (A) 300 pseudo-random volume-pairs were generated, then box filters were applied to each volume with six different kernel sizes (3×3×3, 7×7×7, 11×11×11, 15×15×15, 19×19×19, 23×23×23). Then, the global Moran’s I of 1800 volume-pairs (300 original volume-pairs×6 different box filters) was estimated with varying window (W) size (3×3, 9×9, 15×15, 21×21, 27×27, 33×33, 45×45, 51×51, 57×57). The volume-pairs were then vector-transformed and tested for voxel-by-voxel Pearson’s correlation (topographical correlation). Multiple regression was utilized to test if the global Moran’s I significantly predicted the box-filtering-induced elevation of topographical correlation. The window size of the Moran’s I (W) that gave the best prediction of the topographical correlation from the global Moran’s I was identified using AIC. (B–D) The inflated topographical correlation was observed regardless of the W of Moran’s I while the best prediction resulted when the W of Moran’s I was 51 (lowest AIC).