English

Tail Index Estimation for Discrete Heavy-Tailed Distributions

Statistics Theory 2025-09-23 v3 Statistics Theory

Abstract

It is the purpose of this paper to investigate the issue of estimating the regularity index β>0\beta>0 of a discrete heavy-tailed r.v. SS, \textit{i.e.} a r.v. SS valued in N\mathbb{N}^* such that P(S>n)=L(n)nβ\mathbb{P}(S>n)=L(n)\cdot n^{-\beta} for all n1n\geq 1, where L:R+R+L:\mathbb{R}^*_+\to \mathbb{R}_+ is a slowly varying function. As a first go, we consider the situation where inference is based on independent copies S1,  ,  SnS_1,\; \ldots,\; S_n of the generic variable SS. Just like the popular Hill estimator in the continuous heavy-tail situation, the estimator β^\widehat{\beta} we propose can be derived by means of a suitable reformulation of the regularly varying condition, replacing SS's survivor function by its empirical counterpart. Under mild assumptions, a non-asymptotic bound for the deviation between β^\widehat{\beta} and β\beta is established, as well as limit results (consistency and asymptotic normality). Beyond the i.i.d. case, the inference method proposed is extended to the estimation of the regularity index of a regenerative β\beta-null recurrent Markov chain. Since the parameter β\beta can be then viewed as the tail index of the (regularly varying) distribution of the return time of the chain XX to any (pseudo-) regenerative set, in this case, the estimator is constructed from the successive regeneration times. Because the durations between consecutive regeneration times are asymptotically independent, we can prove that the consistency of the estimator promoted is preserved. In addition to the theoretical analysis carried out, simulation results provide empirical evidence of the relevance of the inference technique proposed.

Keywords

Cite

@article{arxiv.2407.05281,
  title  = {Tail Index Estimation for Discrete Heavy-Tailed Distributions},
  author = {Patrice Bertail and Stephan Clémençon and Carlos Fernández},
  journal= {arXiv preprint arXiv:2407.05281},
  year   = {2025}
}
R2 v1 2026-06-28T17:31:45.313Z