Some Moderate Deviations for Ewens-Pitman Sampling Model

Abstract

Ewens-Pitman model has been successfully applied to various fields including Bayesian statistics. There are four important estimators $K_n,M_{l,n}$,$K_m^{(n)},M_{l,m}^{(n)}$. In particular, $M_{1,n},M_{1,m}^{(n)}$ are related to discovery probability. Their asymptotic behavior, such as large deviation principle, has already been discussed in [4],[1] and [2]. Moderate deviation principle is also discussed in [3] with some speed restriction. In this article, we will apply complex asymptotic analysis to show that this speed restriction is unnecessary.

21 pages, three figures