Distribution of the mean of Brownian motions

24 Views Asked by At

Consider $Z_n = \sum_{k=1}^{n} \frac{B_k}{n}$ where each $B_k \sim N(0, k)$ is a Brownian motion. I'm trying to compute the distribution of $Z_n$. Obviously $EZ_n=0$. For $Var Z_n$, since $\mathbf{Cov}B_sB_t = \min\{s,t\},$ I derive \begin{align} Var Z_n &= \frac{1}{n^2}\bigg(\sum_kVar B_k + 2\sum_{i<j}\mathbf{Cov} B_i, B_j\bigg) \\ &= \frac{1}{n^2}\bigg(\binom{n}{2} + 1(n-1) + 2(n-2) + \ldots+(n-1)\bigg) \\ &= \frac{1}{n^2}\bigg(\binom{n}{2} + \binom{n+1}{3}\bigg) \\ &= \frac{1}{2}\bigg(1-\frac{1}{n}\bigg) + \frac{n}{6}\bigg(1-\frac{1}{n^2}\bigg) \end{align} So $Z_n \sim \mathbf{N\bigg(0, \frac{1}{2}\bigg(1-\frac{1}{n}\bigg) + \frac{n}{6}\bigg(1-\frac{1}{n^2}\bigg)\bigg)}$

There are two things that are a bit confusing: I never used the fact that $B_t - B_s \sim N(0, t-s)$ and the sum of dependent normal rvs is not always normal. I'm not sure how to use these facts though.

1

There are 1 best solutions below

2
On BEST ANSWER

Linear combinations of jointly normal random variables are always normal. Once you use the fact that $EB_tB_s=\min \{t,s\}$ you don't need the fact that $B_t-B_s \sim N(0,t-s)$.