PPT Slide
How do you calculate a multidimensional SD for a skewed
distribution? One way is the BEST method (Bootstrap Error-adjusted Single-sample Technique). Like all statistics, you have to start with an observed data set (calibration set or training set).
Suppose you start with an observed data set T, and you would like to use T to estimate the variability of the population from which it was drawn. The bootstrap lets you make the estimate.