Extremes of Shepp statistics for Gaussian random walk (Q626270)

A random walk with i.i.d. standard Gaussian steps \(\xi_i\) is considered. The author investigates probabilities of large and moderate deviations for the normalized Shepp statistics \[ M_T^{(N)}=N^{-1/2}\max_{1\leq k\leq TN}\max_{1\leq L\leq N} \sum_{i=k}^{k+L-1}\xi_i. \] It is shown that \(P(M_T^{(N)}>u)\sim TN(1-\Phi(u))\) as \(u\to\infty\), \(N\to\infty\) if \[ u/\sqrt{N}\to\infty,\;TN(1-\Phi(u))\to 0,\;\text{and}\;TN\geq 1. \] The limit Gumbel distribution of \(M_T^{(N)}\) with suitable normalization is obtained.