Extreme-Value Analysis of Standardized Gaussian Increments

Abstract: Let

X_{i}, i = 1, 2, ...

be i.i.d. standard gaussian variables. Let

S_{n} = X_{1} + ... + X_{n}

be the sequence of partial sums and

L_n=max_{0leq i<jleq n}frac{S_j-S_i}{sqrt{j-i}}.

We show that the distribution of

L_{n}

, appropriately normalized, converges as

n o i n f t y

to the Gumbel distribution. In some sense, the the random variable

L_{n}

, being the maximum of

n (n + 1) / 2

dependent standard gaussian variables, behaves like the maximum of

H n l o g n

independent standard gaussian variables. Here,

H i n (0, i n f t y)

is some constant. We also prove a version of the above result for the Brownian motion.

This page was built for publication: Extreme-Value Analysis of Standardized Gaussian Increments