Asymptotic expansions in non-central limit theorems for quadratic forms (Q2576799)

Let \(X_i, i\geq 1,\) be a sequence of iid real random variables with (common) distribution function \(F\) and with expectation \(\mathbb{E}X_i =0\). Let \(A:= (a_{i,j})_{1\leq i,j \leq N}\) be a symmetric matrix with eigenvalues \(| \lambda_1| \geq | \lambda_2| \geq \dots | \lambda_N| \). Let further \(Q\) denote the (random) quadratic form \(Q = Q(F,A) := \sum_{j,k=1}^N a_{j,k}(X_jX_k - \mathbb{E}(X_jX_k))\). The authors investigate the limit behaviour in the situation \(D^2 := \sum_j a_{j,j}^2 =0\), hence \(Q = \sum_{j\neq k} a_{j,k}X_j X_k\). In that case the limit distribution is independent from the initial distribution \(F\) [see \textit{V. I. Rotar}', Theory Probab. Appl. 18, 499--507 (1973), translation from Teor. Veroyatn. Primen. 18, 527--534 (1973; Zbl 0304.60037); \textit{V. I. Rotar}' and \textit{T. L. Shervashidze}, ibid. 30, 585--596 (1986), resp. ibid. 30, 549--554 (1985; Zbl 0578.60020); and also \textit{B. A. Sevast'yanov}, ibid. 6, 337--340 (1961), resp. ibid. 6, 368--372 (1961; Zbl 0286.60028)]. Let \(\| A\| \) denote the spectral norm, \(\| A\| ^2 = \sum_{i,j}a_{i,j}^2 = \sum_j \lambda_j^2\). Let \(\mu_k\) denote the \(k\)th moment of \(X_i\). Assuming finite forth moments, the main result, Theorem 1.1, yields uniform estimates of the difference of the distribution functions of \(Q(F,A)\) (normalized by \(\mu_2^{-1}\)) and \(Q(\Phi, A)\), where \(\Phi\) denotes the distribution function of a standard normal distribution. In a particular case, if \(| a_{i,j}| <b/N\), the upper bound is given by \(\frac{C(b)}{N} \mu_4\mu_2^{-1}\), where \(C(b)\) is some constant. The proof consists of a series of technical lemmas, too technical for a review. For the background the reader is referred to the literature, e.g. the authors [J. Theor. Probab. 15, 423--475 (2002; Zbl 1002.60018)], and the literature mentioned there.