Generalized <i>L</i>‐statistics: Correction to the Form of the Asymptotic Variance Presented by Helmers <i>et al.</i> (1990) (Q4956048)

Generalized \(L\)-statistics can be thought of either as weighted \(U\)-statistics or as \(L\)-statistics in which the random variables are not independent. The asymptotic properties of generalized \(L\)-statistics were initially investigated by \textit{B.W. Silverman} [Ann. Probab. 11, 745-751 (1983; Zbl 0514.60040)] and by \textit{R.J. Serfling} [Ann. Stat. 12, 76-86 (1984; Zbl 0538.62015)]. Subsequently, \textit{R. Helmers}, \textit{P. Janssen} and \textit{R. Serfling} [Scand. J. Stat. 17, No. 1, 65-77 (1990; Zbl 0706.62016)] presented a proof indicating that the estimated asymptotic variance of generalized \(L\)-statistics is strongly convergent. The authors of this paper demonstrate that the form of the asymptotic variance used by Helmers, Janssen and Serfling is incorrect. It is shown that this error does not affect the main conclusion, i.e. that the estimated asymptotic variance formed from the empirical distribution function is strongly consistent for the true asymptotic variance.