{"id":"arxiv/2603.19800","url":"https://arxiv.org/abs/2603.19800","tags":["math.PR"],"score":0,"title":"The logarithmic law of sample correlation matrices","domain":"math.PR","source":"arxiv","authors":["Yanpeng Li","Zhi Liu","Jiahui Xie","Wang Zhou"],"reviews":0,"subject":"math.PR","summary":"Let $\\mathbf{R}$ be the sample correlation matrix constructed from $\\mathbf{X}\\in \\mathbb{R}^{p\\times n}$, whose entries are independent and identically distributed random variables with mean zero and tail probability condition $\\lim_{x\\rightarrow \\infty}x^3\\mathbb{P}(|ξ|>x)=0$. We derive the universal logarithmic law for $\\log \\det \\mathbf{R}$, \\begin{equation*} \\frac{\\log \\det \\mathbf{R}-(p-n+1/2)\\log (1-\\frac{p-1}{n})+p-\\frac{p}{n}}{\\sqrt{-2\\log (1-\\frac{p-1}{n})-2\\frac{p}{n}}}\\stackrel{d}{\\rightarrow} {N}(0,1), \\end{equation*} if $p\\le n$ as $p,n\\rightarrow \\infty$. Moreover, under the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-20","reactions":{}}