One-sided refinements of the strong law of large numbers and the Glivenko-Cantelli theorem (Q1201169)

If \(X_ 1,X_ 2,\dots,X_ n\) are i.i.d., then for \(\alpha>2\), the refined (upper) empirical distribution \(G^ \alpha_ n\) is the random probability measure \[ G^ \alpha_ n=\sum^ n_{i=1}\left({1\over n}- {n+1\over 2n^ \alpha}+{i\over n^ \alpha}\right)\delta(X_{i:n}), \] where \(\delta(x)\) denotes the Dirac measure, \(X_{i:n}\) is the \(i\)th order statistic associated with \(X_ 1,X_ 2,\dots,\) \(X_ n\). The Glivenko-Cantelli theorem is proved for \(G^ \alpha_ n\). It is also proved the property that all its quantiles converge almost surely. In some respect, as the authors say, \(G^ \alpha_ n\) is more convenient than the classical empirical distribution function. The proof is based on the investigation of the strong law of large numbers for linear combinations of order statistics.