Table 3.

Goodness of Fit of the Linear, Quadratic, and Exponential Fits to Correlation/Partial Correlation Versus Cortical Distance Plot

	p-Value of correlation between coefficients and distance	BIC of linear fit to mean	BIC of quadratic fit to mean	BIC of exponential fit to mean	% confidence that the model with the best BIC is better than the other models: 100% × (1-Bayes factor)
Corr
Within quadrant	4.28E−49	−1228.52	−1224.77	−1224.63	84.68 Lin > quad	85.69 Lin > exp
Between quadrants	5.05E−24	−1187.35	−1188.31	−1188.43	41.76 Exp > lin	5.85 Exp > quad
Pcorr
Within quadrant	9.33E−36	−1086.76	−1218.09	−1261.53	100 Exp > lin	99.99999996 Exp > quad
Between quadrants	9.37E−05	−1487.55	−1534.07	−1526.42	99.999999992 Quad > lin	97.81 Quad > exp

All columns quantify the data plotted in Figure 9. For each connection, we computed the mean cortical distance across hemispheres and mean correlation/partial correlation coefficient across runs, hemispheres, and subjects. Thus, for within a quadrant, we considered 210 connections: two sets (dorsal and ventral) of 105 coefficients each (number of within-quadrant connections), and for between quadrants, there are 225 coefficients (number of between-quadrant connections). The first column presents the p-value of testing the statistical significance of correlation between functional connectivity measures and cortical distances. Goodness of fit (columns 2–4) was evaluated using the BIC. A smaller BIC (more negative) implies a better fit relative to the other models. Bold values indicate the best BIC. To test differences between measures of goodness of fit, a Bayes factor was computed. We present the percent confidence that one model is better than the others tested (columns 5–6). Bold font indicate a strong confidence that one model is a better choice than the others.

BIC, Bayesian Information Criterion.