where i k, x is k-th sample of i-th variable Xi, and
1 Fi
− is the inverse CDF for the variable Xi. The objection
against the approach deals mainly with the
tails of PDF, which mostly influences the variance,
skewness and kurtosis of the sample set. However
this elementary approach was overcome by sampling
the mean values related to the intervals (e.g. Huntington
& Lyrintzis, 1998), see Fig. 1: