| Abstrakt: |
A new formula is derived for the random error of sample central moments from correlated data which does not assume an underlying distribution and is accurate to leading order in the number of sample elements. Central moments, being important quantities in turbulence research, require accurate error estimation. Many approaches have been followed in the past for estimating the random errors of central moments from correlated data. These include: simple extensions of the formula for independent data, using the formula for the random error of generic averages, assuming an underlying normal distribution, and using block bootstraps. All of these approaches are compared with the present formula using datasets from a turbulent boundary layer, freestream grid turbulence, and a turbulent round jet. For even-order sample central moments, many of the existing approaches perform well with differences of <15%. However, for odd-order sample central moments, only the block bootstrap methodology performs similarly well. For the same sample central moments, the other methods differ by as much as 200–1000%. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink