On the multiplicity of the sample maximum and the longest head run (Q5933710)

Several interesting results are proved in this article. Here is an extremly abbreviated description. Let \(Y_i\) be i.i.d. nonnegative integer valued random variables, satisfying the property of `stabilizing hazard', that is Pr\(\{Y_1=k\}\)/Pr\(\{Y\geq k\}\rightarrow r\in (0,1)\). Then, if \(\mu (n)\) is the multiplicity of the maximal value among the first \(n\) variables, then \(\{\mu (n)\}\) is stochastically bounded, but does not converge in distribution in general, however, it has an a.s. limit distribution in the sense that \[ \lim_{t\to\infty}(\log t)^{-1}\sum\limits_{n=1}^{t} n^{-1}I(\mu(n)=m)=r^m/(m\log(1-r)^{-1}). \] A similar result holds for the distribution of a generalization of the multiplicity of the longest head run.