A METHOD TO DETERMINE FRACTILE VALUES FROM STATISTICAL DATA
Toshiyuki Sugiyama, Yozo Fujino, Manabu Ito · Doboku Gakkai Ronbunshu · 1985
A method to determine fractile values from statistical data is proposed: the fractile value for an exceedance probability about 0.5 is first determined from the data and next, the correction factor which is also calculated from the data is multiplied to that fractile, yielding the fractile value of a required exceedance probability. By means of Monte Carlo simulation, its accuracy and those of commonly used methods are compared. The result shows that proposed method is the most stable to various distribution models and coefficients of variation.