The normalization is standard score (z-score) normalization applied for each variable over all samples. I.e. assuming we have N samples, then for each variable let { x_i, i = 0...N-1 } be the corresponding variable values over the samples and
m = mean(x) = SUM(x_i / N)
s = std(x) = sqrt( SUM( (x_i - m)^2 / (N - 1) ) )
The normalized variable values y_i are then computed as
y_i = (x_i - m) / s |
