{"id":"arxiv/2602.17960","url":"https://arxiv.org/abs/2602.17960","tags":["math.PR","stat.AP","stat.ML"],"score":0,"title":"Anisotropic local law for non-separable sample covariance matrices","domain":"math.PR","source":"arxiv","authors":["Zhou Fan","Renyuan Ma","Elliot Paquette","Zhichao Wang"],"reviews":0,"subject":"math.PR","summary":"We establish local laws for sample covariance matrices $K = N^{-1}\\sum_{i=1}^N \\g_i\\g_i^*$ where the random vectors $\\g_1, \\ldots, \\g_N \\in \\R^n$ are independent with common covariance $Σ$. Previous work has largely focused on the separable model $\\g = Σ^{1/2}\\w$ with $\\w$ having independent entries, but this structure is rarely present in statistical applications involving dependent or nonlinearly transformed data. Under a concentration assumption for quadratic forms $\\g^*A\\g$, we prove an optimal averaged local law showing that the Stieltjes transform of $K$ converges to its deterministic…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-02-20","reactions":{}}