On the error bound in a combinatorial central limit theorem (Q2345123)

Let \(\mathbb{X}=\{ X_{ij}: 1 \leq i,j \leq n \} \) be an \(n \times n\) array of independent random variables where \(n \geq 2\), \(\pi\) a uniform random permutation of \(\{ 1,2,\cdots,n\}\), independent of \(\mathbb{X}\), and \( W =\sum_{i=1}^n X_{i\pi(i)}\). Suppose \(\mathbb{X}\) is standardized so that \({\mathrm E} W=0\), \(\mathrm{Var}(W)=1\). Using Stein's method of exchangeable pairs and a concentration inequality, this paper proves that the Kolmogorov distance between the distribution of \(W\) and the standard normal distribution is bounded by \(451 \sum_{i,j=1}^n {\mathrm E} |X_{ij}|^3/n\).