plotregression actually uses the regression function (also from Neural Network Toolbox) to compute the R value.
However, note that results are almost identical to corrcoef's output.
load simplefit_dataset
net = newff(simplefitInputs,simplefitTargets,10);
[net,tr] = train(net,simplefitInputs,simplefitTargets);
simplefitOutputs = sim(net,simplefitInputs);
plotregression(simplefitTargets,simplefitOutputs);
R value in plotregression is computed this way:
[r,m,b] = regression(simplefitTargets,simplefitOutputs);
Using corrcoef:
c = corrcoef(simplefitTargets,simplefitOutputs);
Values should be almost identical. In my test:
norm(c(1,2) - r)
ans =
5.551115123125783e-16
