Extremal processes, secretary problems and the 1/e law

From MaRDI portal
Publication:3468385





The basic assumption in most versions of the classical ``Secretary Problem is that the order of arrival of the n candidates to be interviewed is totally random, assigning an equal probability of 1/n! to each of the possible permutations of the candidates' ranks. It follows that, if \(X_ i\) is the rank of the i-th candidate to appear, then \(\{X_ i\}\), \(i=1,...,n\), is a sequence of independent random variables, each of which is uniformly distributed on the integers 1,...,n. In the problem considered here, \(X_ 1,...,X_ n\) are still assumed to be independent, but the distribution of \(X_ i\) is assumed to be of the form \(F^{\alpha_ i} (\alpha_ i>0)\). This extension allows for situations in which the best candidates are clustered either at the beginning (the sequence \(\{\alpha_ i\}\) is decreasing) or at the end (the sequence \(\{\alpha_ i\}\) is increasing) of the arrival process. Let \(p_ i=\alpha_ i/\sum^{i}_{k=1}\alpha_ k\); the main result of the paper states that, assuming \(\sum^{\infty}_{i=1}p_ i=\infty\) and \(\sum^{\infty}_{i=1}p_ i^ 2<\infty\), then the probability of selecting the best candidate is asymptotically equal to 1/e, as in the classical version of the problem.




Cited in
(30)








This page was built for publication: Extremal processes, secretary problems and the 1/e law

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3468385)