Figure 5.

Dependence of the best fit ${\chi }_{\mathrm{red}}^2$ on the number of neighbors included in the fit. Panel (a) shows ${\chi }_{\mathrm{red}}^2$ versus neighbor number n, and panel (b) shows these fitted values in a colormap. Beyond the eighth neighbor, the ${\chi }_{\mathrm{red}}^2$ does not appreciably improve by adding additional neighbors, so we truncate our model at the eighth neighbor.