I want to use the Latin hypercube sample function (lhsdesign). In the examples provided in the documentation the quality of the generated design is measured by calculating the sum of between-column aquared correlations,
rng default % For reproducibility
X= lhsdesign(10,4); % X is a 10-by-4 design matrix.
(sum(y(:).^2)-4)/2 % the 4 subtracted to remove the ones on the diagonal of y.
Why was the sum of the squared correlations divided by 2 ?