# included for speed. it is faster than upper.lower
total.multicorrelation.bias.corrected.upper( x, vec = 1:ncol(x), squared = FALSE, ... )