Normal approximation for associated point processes via Stein's method with applications to determinantal point processes (Q2325926)

Let \(X\) be a (positively or negatively) associated point process on \(\mathbb{R}^d\). Under some relatively mild boundedness, and similar, conditions, the author establishes an explicit bound in Wasserstein (or \(L^1\)) distance in the central limit theorem for functionals of \(X\). More specifically, letting \(f_{\mathbf{i}}\) be real-valued, measurable functions for \(\mathbf{i}\in\mathbb{Z}^d\), define \[ S_{\mathbf{k}}^n=\sum_{\mathbf{i}\in B_{\mathbf{k}}^n}[f_{\mathbf{i}}(X\cap C_{\mathbf{i}})-\mathbb{E}f_{\mathbf{i}}(X\cap C_{\mathbf{i}})]\,, \] where \(C_{\mathbf{i}}\) is the \(d\)-dimensional unit cube centred at \(\mathbf{i}\), \[ B_{\mathbf{k}}^n=\left\{\mathbf{i}\in\mathbb{Z}^d : \mathbf{k} \leq \mathbf{i} < \mathbf{k} + n\mathbf{1} \right\}\,, \] \(\mathbf{1}=(1,1,\ldots,1)\), and where inequalities are interpreted componentwise. The main result of the present paper is an explicit bound, with rate \(n^{-d/(4d+2)}\), on the Wasserstein distance between \(S_{\mathbf{k}}^n/\mbox{Var}(S_{\mathbf{k}}^n)\) and a standard Gaussian random variable. This result is applied to determinantal point processes, which are known to be negatively associated, with special attention given to the Laguerre-Gaussian family of determinantal point processes. The proof of the main result uses Stein's method for Gaussian approximation, and exploits connections between association and \(\alpha\)-mixing.