2
$\begingroup$

Let $(X_i)_i$ be independent random variables such that $X_i \sim U[0, 1]$, and define $Y_i = X_i X_{i+1}$ and $Z_n = \sum_{i=1}^n Y_i$.

I want to find the variance of $Z_n$.

I found that $\mathbb E [Z_n]=\frac{n}{4}$ , and when applying the formula of the variance, I have $\sum_{i=1}^n \operatorname{Var}(Y_i) = \frac{7n}{144}$ using the fact the $X_i$'s are independent.

How do I find the covariance of the other variables?

$\endgroup$
2
$\begingroup$

$$\text{Var}\left(\sum_{i=1}^n Y_i \right) = \sum_{i=1}^n \sum_{j=1}^n \text{Cov}(Y_i, Y_j)$$

Now if $|i-j|\ge 2$, $Y_i = X_i X_{i+1}$ and $Y_j = X_j X_{j+1}$ are independent, so these covariances are $0$.

If $i=j$, $$\text{Cov}(Y_i, Y_i) = \text{Var}(X_i X_{i+1}) = \mathbb E[X_i^2 X_{i+1}^2] - \mathbb E[X_i X_{i+1}]^2 = \mathbb E[X_i^2]^2 - \mathbb E[X_i]^4 =\frac{7}{144}$$ There are $n$ of these terms, contributing a total of $7n/144$ to the variance of $Z_n$.

If $i=j\pm 1$, $$\eqalign{\text{Cov}(Y_i,Y_j) &= \text{Cov}(X_{j+1} X_{j+2},X_{j} X_{j+1}) = \mathbb E[X_j X_{j+1}^2 X_{j+2}] - E[X_j X_{j+1}]^2 \cr &= \mathbb E[X_j]^2 \mathbb E[X_j^2] - \mathbb E[X_j]^4 = \frac{1}{48} }$$

There are $2n-2$ of these terms, contributing a total of $(2n-2)/48$. Thus $$ \text{Var}(Z_n) = \frac{7n}{144} + \frac{2n-2}{48} = \frac{13 n}{144} - \frac{1}{24} $$

$\endgroup$
2
$\begingroup$

You have $$ \operatorname{Var} Z_n = \operatorname{Var} \sum_{k=1}^n Y_k $$ but sadly you cannot write that this is equal to $\sum_{k=1}^n \operatorname{Var} Y_k$ as the $Y_k$'s are not independent (and not even pairwise independent either, which would have been sufficient).

However, you can write: $$ \mathbb{E} Z_n^2 = \mathbb{E}\left[ \left(\sum_{k=1}^n Y_k \right)^2 \right] = \mathbb{E}\left[ \sum_{k=1}^n \sum_{\ell=1}^n Y_kY_\ell \right] = \sum_{k=1}^n \sum_{\ell=1}^n \mathbb{E}\left[Y_kY_\ell \right] = \sum_{k=1}^n \sum_{\ell=1}^n \mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right]\,. $$ Now, let us focus on a term $\mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right]$.

  • if $k=\ell-1$, then $$ \mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right] = \mathbb{E}\left[X_kX_\ell^2 X_{\ell+1} \right] = \mathbb{E}\left[X_k\right]\mathbb{E}\left[X_\ell^2\right]\mathbb{E}\left[ X_{\ell+1} \right] = \frac{1}{2}\cdot \frac{1}{3} \cdot\frac{1}{2} = \frac{1}{12} $$

  • if $k=\ell$, then $$ \mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right] = \mathbb{E}\left[X_k^2 X_{k+1}^2 \right] = \mathbb{E}\left[X_k^2\right]\mathbb{E}\left[X_{k+1}^2\right] = \frac{1}{3}\cdot \frac{1}{3} = \frac{1}{9} $$

  • if $k=\ell+1$, then $$ \mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right] = \mathbb{E}\left[X_k^2 X_\ell X_{k+1} \right] = \mathbb{E}\left[X_k^2\right]\mathbb{E}\left[X_\ell\right]\mathbb{E}\left[ X_{k+1} \right] = \frac{1}{3}\cdot \frac{1}{2} \cdot\frac{1}{2} = \frac{1}{12} $$

  • else, $$ \mathbb{E}\left[X_kX_{k+1}X_\ell X_{\ell+1} \right] = \mathbb{E}\left[X_k\right]\mathbb{E}\left[ X_{k+1} \right]\mathbb{E}\left[X_\ell\right]\mathbb{E}\left[ X_{\ell+1} \right] = \frac{1}{16} $$

Overall, $$ \mathbb{E} Z_n^2 = \sum_{k=1}^n \frac{1}{9} + 2\sum_{k=1}^{n-1} \frac{1}{12} + \sum_{k=1}^{n}\sum_{\ell=1}^{n} \frac{1}{16}\mathbf{1}_{|k-\ell|>1} = \frac{n}{9}+\frac{n-1}{6} + \frac{n^2-n-2(n-1)}{16} = \frac{9n^2+13n-6}{144} $$ and thus $$ \operatorname{Var} Z_n = \mathbb{E} Z_n^2 - \mathbb{E}[ Z_n ]^2 = \frac{9n^2+13n-6}{144} - \frac{n^2}{16} = \boxed{\frac{13n-6}{144}} $$


As a sanity check, for $n=1$ we want the variance of $XX'$ where $X,X'$ are independent r.v.'s uniform in $[0,1]$; it is not hard to check that the variance in this case is $\frac{7}{144}$, matching the above result.

$\endgroup$

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.