2 SAMPLING AND STATISTICAL
CORRELATION
There are two stages to the Latin Hypercube Sampling.
First, samples for each (marginal) variable are
chosen strategically to represent the variable’s probability
density function PDF. The NSim samples
(where NSim is the number of the planned simulations)
for each random variable Xi are often chosen
from the cumulative distribution function (CDF) by
the inverse transformation of CDF.