skekurx(3f) - [M_math:statistics] Compute unbiased estimator of the population SKEWNESS and KURTOSIS
Synopsis
Description
Options
Returns
Examples
Author
SUBROUTINE SKEKURX(Y,N,YSKEW,YKURT)
integer,intent(in) :: n real,intent(in) :: y(*) real,intent(out) :: yskew real,intent(out) :: ykurt
This routine calculates the unbiased estimator of the population kurtosis and skewness from a subset of samples
kurt = {n*(n+1)/((n-1)*(n-2)*(n-3))*SUM[((x(i)-xbar)/stddev)**4]} - 3*(n-1)**2/((n-2)*(n-3))where xbar and stddev are the sample mean and standard deviationskew = ( n / ((n-1)*(n-2)) *SUM{((X(i) - xbar)/stddev)**3}
Note that this is apparently the skewness and kurtosis calculated by the MicroSoft Excel product. I checked the Excel help and Excel uses the above formulas. No references are given in the Excel documentation. Note that this converges on the standard expression for excess kurtosis and skewness as N becomes large.
John S. Urban
Nemo Release 3.1 | skekurx (3) | February 23, 2025 |