1
$\begingroup$

Suppose $X_n\sim N(0,1) $ is iid, then it is easy to see that $$\sum_{n=1}^{\infty}\frac{X_n}{n}\cos nx$$ converges a.s. for any $x$ since $$\sum_{n=1}^{\infty}var(\frac{X_n}{n}\cos nx)<\infty$$ but how to show that the convergence is uniform? That is, for a mesurable set $A\subset \Omega, \mathbb{P}(A)=1$, the series converges for all $x\in\mathbb{R}$ and $\omega\in A$.

$\endgroup$
2
  • 2
    $\begingroup$ Can you explain what uniform convergence is in this case. You have a function of $\omega$ from the underlying probability space and $x$. $\endgroup$ May 30, 2023 at 11:58
  • 1
    $\begingroup$ For a $A\subset \Omega, \mathbb{P}(A)=1$, the series converges for all $x\in\mathbb{R}$ and $\omega\in A$. $\endgroup$
    – Sheng Wang
    May 30, 2023 at 13:07

1 Answer 1

2
$\begingroup$

Let \begin{equation*} F(x):=\sum_{n=1}^{\infty}\frac{X_n}n\,\cos nx. \tag{1}\label{1} \end{equation*} For $j=0,1,\dots$, let \begin{equation*} s_j:=\sqrt{\sum_{2^j\le n<2^{j+1}}E\Big(\frac{X_n}n\Big)^2}=\sqrt{\sum_{2^j\le n<2^{j+1}}\frac1{n^2}}. \end{equation*}

For any integer $N>0$ and
\begin{equation*} S_N:=\sum_{N\le n<2N}\frac1{n^2} \end{equation*} we have $S_{N+1}-S_N=-\dfrac1{N^2}+\dfrac1{(2N)^2}+\dfrac1{(2N+1)^2}<0$. So, $S_N$ is decreasing in $N=1,2,\dots$ and hence $s_j$ is decreasing in $j=0,1,\dots$. Also, $s_j^2\asymp2^j\frac1{(2^j)^2}\to0$ as $j\to\infty$.

So, by Theorem 1, p. 84 in Kahane's book, the function $F$ is continuous almost surely (a.s.). (The condition in that theorem that $s_j$ be decreasing seems possible to relax.) As noted on p. 48 of Kahane's book, the a.s. continuity of the sum of a random Fourier series of the form \eqref{1} is equivalent to the a.s. uniform convergence of the random series. $\quad\Box$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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