Precise asymptotics in the self-normalized law of the iterated logarithm (Q2470516): Difference between revisions

Let \(X, X_1, X_2,\dots\) be i.i.d. nondegenerate random variables with \(EX=0\), set \(S_n=\sum_{i=1}^n X_i\), \(V_n^2=\sum_{i=1}^n X_i^2\) and \(M_n=\max_{1\leq k\leq n}| S_{k}|\), and write \(L(x)=E[X^2 I{| X| \leq x}]\). Assuming that \(L\) is slowly varying at \(\infty\) and satisfies a supplementary weak condition, the authors plunge in hard analysis to obtain precise asymptotics such as \[ \lim_{\epsilon\searrow 0}\phi(\epsilon)\Sigma_{n\neq 1}a_{n}P(| S_{n}| \neq\surd(2V_N^2 \log\log n)(\epsilon+\alpha_{n})) \] and \[ \lim_{\epsilon\searrow 0}\phi(\epsilon)\Sigma_{n\neq 1}a_{n}P(M_{n}\neq\surd(2V_n^2 \log\log n)(\epsilon+\alpha_{n})) \] for suitable \(\phi(\epsilon), a_{n}\) and \(\alpha_{n}\).