Key points are not available for this paper at this time.
We ask: Which frequency shall we attribute to the m'th observation of a statistical variate, and especially to the m'th flood‐discharge? Let x m m = 1, 2,...... (n − 1), n be n observations arranged according to increasing magnitude. The question is, shall we use (m/n) or (m − 1 ) / n as the observed cumulative frequency? Both methods are equally legitimate. The first corresponds to the arrangement of the observations in increasing magnitude, the second to the arrangement in decreasing magnitude.
E. J. Gumbel (Fri,) studied this question.