A weak invariance principle for weighted \(U\)-statistics with varying kernels (Q689368): Difference between revisions

Let \(\{X_ n\}\) be a sequence of i.i.d. random variables uniformly distributed on \((0.1)\). Consider a sequence of weighted \(U\)-statistics of the form \(U_ n = \sum^ n_{i,j = 1} a^{(n)}_{ij}h_ n(X_ i,X_ j)\), \(n = 1,2,\dots\), where \(Eh^ 2_ n(X_ 1,X_ 2) = 1\), \(E(h_ n(X_ 1,X_ 2)\mid X_ 2) = 0\) a.s. and \(\sum_{ij}(a^{(n)}_{ij})^ 2 = 1/2\) for every \(n\). Let \(\{u_ n\}\) be a sequence of random processes with continuous paths defined by \(U_ n\) and let \(\{v_ n\}\) be a sequence of continuous Gaussian processes suitable constructed from i.i.d. standard normal random variables. It is proved that, under some additional assumptions, the Lévy-Prokhorov distance between \(u_ n\) and \(v_ n\) converges to zero. A number of weak limit theorems of functional type are derived from this general result.