I am checking the proof of the construction of a brownian motion in $[0,\pi]$. We show that
\begin{gather*} t \mapsto B^m_t = \frac{t}{\sqrt{\pi}}X_0 + \sqrt{\frac{2}{\pi}}\sum_{n=1}^{2^m-1}X_n \frac{\sin(nt)}{n}, \end{gather*} where $(X_n,n\in\mathbb{N})$ is a sequence of i.i.d. random variables $N(0,1)$, converges uniformly on $[0,\pi]$ with probability $1$ to a continuous function, denoted $t \mapsto B_t$, that is a standard brownian motion.
When we have proved this, it remains to show that \begin{gather*} B_t = \frac{t}{\sqrt{\pi}}X_0 + \sqrt{\frac{2}{\pi}}\sum_{n=1}^{+\infty}X_n \frac{\sin(nt)}{n} \end{gather*} is indeed a brownian motion. A sentence in this proof says that "For $s,t \in \mathbb{R}_+$, $B_{s+t} - B_s$ is gaussian as a sum of independent gaussian random variables". Why is it true for series ?


Define for fixed $s,t\in\mathbb R_+$ the random variable $$Y_N:=\frac{t}{\sqrt{\pi}}X_0+\sqrt{\frac 2{\pi}}\sum_{n=1}^NX_n\frac{\sin(nt)-\sin(ns)}n.$$ For a fixed $N$, the random variable $Y_N$ has Gaussian distribution has a linear combination of the Gaussian vector $(X_0,\dots,X_N)$.

By the results mentioned in the OP, the sequence $(Y_N)_{N\geqslant 1}$ is converges in distribution (actually almost surely), and the limit in distribution of a Gaussian distribution is Gaussian.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.