A nonparametric measure of independence under a hypothesis of independent components (Q1200738): Difference between revisions

It is shown that, under certain regularity assumptions, the weighted \(L^ 2\)-distance \[ I_ n=\iint\bigl[f_ n(x_ 1,x_ 2)-g_ n(x_ 1)h_ n(x_ 2)\bigr]^ 2 a(x_ 1,x_ 2) dx_ 1 dx_ 2 \] between a bivariate kernel density estimator \(f_ n\) and the tensor product of its marginals \(g_ n\) and \(h_ n\) is asymptotically normal after a proper standardization, provided that the components of the observed (bivariate) data are independent. The method of proof utilizes a central limit theorem for degenerate \(U\)- statistics due to \textit{P. Hall} [J. Multivariate Anal. 14, 1-16 (1984; Zbl 0528.62028)].