Hölderian invariance principle for triangular arrays of random variables (Q1881770)

Let \(w_\alpha(x,\delta) = \sup_{0<| t-s| <\delta} | x(t) - x(s)|/| t - s|^\alpha\). Let \({\mathcal H}^0_\alpha\) \((0 < \alpha < 2^{-1})\) be the Hölder space of functions \(x : [0,1]\to \mathbb{R}\), equipped with the norm \(\| x\|_\alpha = | x(0)| + w_\alpha(x,1)\). Let \(\{X_{n,k}\); \(k =1,\dots,k_n\}\) be triangular arrays of row-wise independent mean zero random variables. Assume \(\sigma^2_{n,j}= EX^2_{n,j} < \infty\) and define the process \(\{\Xi_n(t): t\in [0,1]\}\) by \[ \Xi_n(t) =\sum^{k-1}_{j=1} X_{n,j} +\frac{t-b_n(k)}{\sigma^2_{n,k}} X_{n,k}(b_n(k-1)\leq t\leq b_n(k))\quad (k= 1,\dots,k_n), \] where \(b_n(k) = \sigma^2_{n,1}+\dots+\sigma^2_{n,k}\). The authors prove, among others, that if for every \(\varepsilon > 0\) \[ \lim_{n\to\infty}\sum^{k_n}_{k=1} P(| X_{n,k}|\geq\varepsilon\sigma^{2\alpha}_{n,k})= 0,\qquad \lim_{n\to\infty} \sum^{k_n}_{k=1}EX^2_{n,k}I(| X_{n,k}|\geq\varepsilon)=0 \] and for some \(q > (2^{-1}-\alpha)^{-1}\) \[ \lim_{\tau\to 0}\limsup_{n\to\infty}\sum^{k_n}_{k=1}\sigma^{-2\alpha q}_{n,k}EX^q_{n,k}I(| X_{n,k}|\leq \tau\sigma^{2\alpha}_{n,k})=0, \] then as \(n\to\infty\), \(\Xi_n\) converges weakly to a Brownian motion in the space \({\mathcal H}^0_\alpha\). Applying the result, the auhtors obtain necessary and sufficient conditions for the almost sure (resp. in probability) weak Hölder convergence of partial-sum processes based on bootstrapped samples.