Statistics for unimodal sequences (Q2131752)

A sequence \(\lambda\) of positive integers is said to be unimodal if there is a peak \(\lambda_{PK}\) in the sequence: \(\lambda_r^{[L]}\le \dots\le \lambda_1^{[L]}\le \lambda_{PK}\ge \lambda_1^{[R]} \ge \dots \ge \lambda_s^{[R]}\). This unimodal sequence has size (or weight) \(n\) if \(\lambda_{PK}+ \sum_{i=1}^r \lambda_i^{[L]}+\sum_{j=1}^s\lambda_j^{[R]}=n\). If we replace all \(\le\) and \(\ge\) with strict inequalities we arrive at the strong unimodal sequences. In the paper under review, the authors prove a number of limiting distributions for statistics for unimodal sequences. Their Theorem 1.3 is an analogue of the partition theorem of \textit{P. Erdős} and \textit{J. Lehner} [Duke Math J. 8, 335--345 (1941; JFM 67.0126.02, Zbl 0025.10703)] on the largest part of almost all partitions of \(n\) (\(n\to \infty\)) which varies from the mean by an extreme value distribution. Theorems 1.4, 1.5, and 1.11 are analogues of the partition theorems of \textit{B. Fristedt} [Trans. Am. Math. Soc. 337, 703--735 (1993; Zbl 0795.05009)]. According to Theorem 1.8, the total small part counts on the left and the right behave as independent extreme value distributions when the mean is subtracted. Theorem 1.10 shows that Thms. 1.4 and 1.5 have analogues for strongly unimodal sequences. The proofs adapt Fristedt's techniques using conditioned Boltzmann model.