Does $Z_n=\sum_{k=1}^{n}\sqrt{k}X_k$ satisfy the strong law of large numbers if $ X_n: \begin{matrix}-\frac{1}{n} & \frac{1}{n} \\ \frac{1}{2} & \frac{1}{2} \end{matrix}, n=1,2,...$ are independent random variables.

I have the following theorems, but I cannot prove this, I have tried all which I understand, the theorems I have are:

1.) Strong Law of large numbers states that the sequence $X_1,X_2,...$ must satisfy:

$$\frac{1}{n}\sum_{k=1}^{n}(X_k-EX_k)\to^{a.s.}0, n\to \infty$$

2.) Kolmogorov Law: If $(X_n)$ independent random variables, such that $\sum_{n=1}^{\infty} \frac{\text{Var}(X_n)}{n^2}<\infty$, then the strong law of large numbers is satisfied.

3.)Borels: If $ S_n:\mathcal B(n,p)$ (binomial distribution), then $$\frac{S_n}{n}\to^{a.s.}p, n \to \infty$$

or the consequence:

Let $X_n$, sequence of independent random variables, equally distributed, such that $EX_k=a$ and $\text{Var}X_k= \omega^2, k=1,2,3... \implies$

$$\frac{1}{n}\sum_{k=1}^{n}X_k\to^{a.s.}a, n\to \infty$$

  • $\begingroup$ There's no denominator in $\sum\sqrt{k}X_k$? $\endgroup$ – Landon Carter Nov 11 '15 at 18:19
  • $\begingroup$ thats right ... $\endgroup$ – Bozo Vulicevic Nov 11 '15 at 18:19

Consider the random variables $Y_n=\sqrt{n}X_n$. Then note $Y_n$ are independent, and $E(Y_n)=0$. Further note that $\sum_{n=1}^\infty Var(Y_n)=\infty$.

Now invoke Kolmogorov's Three Series Theorem. One of the converging series must be $\sum_nVar(Y_n^c)$ for every $c>0$ if $\sum_nY_n<\infty$ almost surely.

Here, $Y_n^c=Y_n$ if $|Y_n|\leq c$ and $0$ otherwise.

Note that $Y_n=\pm\dfrac{1}{\sqrt{n}}$ each with prob. $1/2$ so for any given $c$, there exists $N\in\mathbb N$ such that $|Y_n|\leq c$ for all $n\geq N$.

Hence for sufficiently large $n$ we must have $Y_n^c=Y_n$.

So $\sum_nVar(Y_n^c)\geq \sum_{n\geq N}Var(Y_n^c)=\sum_{n\geq N}Var(Y_n)=\infty$.

Hence almost sure convergence of $\sum_{n}\sqrt{n}X_n$ does not happen.

| cite | improve this answer | |
  • $\begingroup$ I dont know this three series theorem that you say, Im not allowed to use theorems that we havent studied $\endgroup$ – Bozo Vulicevic Nov 11 '15 at 18:59
  • $\begingroup$ What if you prove the theorem? ;) $\endgroup$ – Landon Carter Nov 11 '15 at 19:07
  • $\begingroup$ But then I would have to state the theroem precisely and prove it using what is known til then, I guess, but how would one go about doing that? ;) $\endgroup$ – Bozo Vulicevic Nov 11 '15 at 19:43
  • $\begingroup$ On a serious note, it is an important tool and I don't know why you haven't been taught it as yet, as it uses no more intricate concept. You can find the theorem and its proof here: math.nus.edu.sg/~matsr/ProbI/Lecture4.pdf . In any case, you don't need to prove the "entire" three series theorem. Just show that if for some $c>0$, $\sum_nVar(Y_n^c)=\infty$ then $\sum_nY_n$ does not converge almost surely. $\endgroup$ – Landon Carter Nov 12 '15 at 4:37

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.