Source: Hedderich, Sachs and Reynarowych, Applied Statistics: Methods Using R,
section 7.2.8, equation (7.33), citing Anderson and Darling (1952) and Stephens
(1986b). The corpus text layer renders (7.33) with two extraction defects, corrected
here against the surrounding prose and Anderson and Darling (1952): the trailing "n
sigma 2" is spillover from the neighbouring column and is not part of the statistic,
and "1 - Y_{N+1-i}" is "1 - F(Y_{N+1-i})". What is computed is the standard
published form A2 = -N - (1/N) sum_i (2i-1) (log u_i + log(1 - u_{N+1-i}))
with u_i the sorted probability-integral transforms.