The use of the Pearson type 3 and log Pearson type 3 distributions revisited

Different methods of fitting the Pearson type 3 and log Pearson type 3 distributions are applied to several long‐term records of annual flood flows previously tested for independence and homogeneity. By using this empirical comparison, it is shown that (1) the Pearson type 3 distribution conforms generally better to annual flood data than the log Pearson type 3 distribution, (2) it is preferable with respect to the log Pearson type 3 distribution to apply the method of moments to the observed sample rather than to the logarithmically transformed sample, and (3) it is important to correct the bias of the coefficient of skewness