The need for reliable simulation systems has been discussed and recognized
by many researchers. At the same time researchers have recognized that the
quality of a simulation model's results are strictly related to the quality
of the input probability distribution functions. The data used in this res
earch were acquired from the Atkinson-Washington-Zachry joint venture on th
e Eastside Reservoir Project in California. The data were analyzed using Be
stFit software to obtain the parameters of the theoretical distribution fun
ctions that best described the field data set. The research validated previ
ous warnings about the influence of the class interval decision on the sele
ction of a distribution function when the chi-square fitting test is utiliz
ed. A second issue of importance that was encountered was the reliability o
f goodness-of-fit tests when dealing with large data sets.