e are interested in calculation of the derivative
evaluated by Monte-Carlo simulation of the
The naive attempt to apply finite
does not work because each of the components is evaluated by the Monte-Carlo
procedure with a random error. These errors do not offset. Dividing it by the
amplifies the error in the final result.