Regularity of the characteristic function of additive functionals for iterated function systems. Statistical applications (Q2866536)

Classical limit theorems of probability theory in various settings are often proved using Fourier techniques. The authors of the present paper add to the vast literature on the subject by completing and improving the results of \textit{H. Hennion} and \textit{L. Hervé} [Ann. Probab. 32, No. 3A, 1934--1984 (2004; Zbl 1062.60017)] and partly of \textit{L. Hervé} and \textit{F. Pène} [Bull. Soc. Math. Fr. 138, No. 3, 415--489 (2010; Zbl 1205.60133)] to general additive functionals of the form \(S_n=\sum_{k=1}^n\xi(\vartheta_k, X_{k-1})\). Here, \((X_n)_{n\in\mathbb{N}}\) is an iterated function system (IFS), i.e., a sequence of \(\mathbb{X}\)-valued r.v.'s, defined recursively via \(X_n=F(\vartheta_n, X_{n-1})\), \(n\geq1\), where \((\vartheta_n)_{n\geq1}\) are \(\mathbb{V}\)-valued i.i.d.\ r.v.'s, independent of \(X_0\) and \(F:\mathbb{V}\times\mathbb{X}\to\mathbb{X}\) is a measurable map such that \(F(\theta,\cdot)\) is Lipschitz with a finite Lipschitz constant \(L_\theta\) for any \(\theta\in\mathbb{V}\), \((\mathbb{X},d)\) is a complete metric space and \(\mathbb{V}\) is a measurable space. The map \(\xi:\mathbb{V}\times\mathbb{X}\to\mathbb{R}^N\) is a weighted-Lipschitz functional.NEWLINENEWLINE The authors present a general condition, which in particularly important cases reduces to simple and (almost) optimal moment conditions, that allows to obtain the main result of the paper, namely the \(\mathcal{C}^m\)-differentiability near \(t=0\) of the characteristic function \(t\mapsto \mathbb{E}\exp\{i\langle t,S_n^\mu\rangle\}\); here \(\mu\) denotes the law of \(X_0\). This result, then provides a way to precisely control the constants in certain limit theorems (e.g., Berry-Esseen theorem); furthermore, it is also applied here to investigate the rate of convergence in the CLT for general \(M\)-estimators associated to an IFS.