Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Suppose $X_1,X_2,\ldots$ are $m$-dependent random variables. Let $F_i$ be the cdf of $X_i$. Let $F_n(x, \omega)$ be the empirical cdf of $X_1,\ldots,X_n$. What will be the variance of $F_n(x, \omega)$?

share|improve this question
    
Do you mean $F_i(x) = \mathbb{P}(X_i \leqslant x)$? Also, for the empirical cdf, do you mean $F_n(x,\omega) = \frac{1}{n} \sum_{k=1}^n I(X_k(\omega) < x)$? It could not hurt to be a little more specific. –  Sasha Apr 30 '12 at 5:11
    
Yes and thank you for the clarification. –  user12847 Apr 30 '12 at 5:42
add comment

1 Answer 1

up vote 0 down vote accepted

Let $Y(\omega) = F_n(x, \omega) = \frac{1}{n} \sum_{k=1}^n I(X_k(\omega) \leqslant x)$. Then, using $\mathbb{E}(I(X_k \leqslant x)) = \mathbb{P}(X_k \leqslant x)$, we have $$ \mathbb{E}\left(Y\right) = \frac{1}{n} \sum_{k=1}^n \mathbb{P}(X_k \leqslant x) = \frac{1}{n} \sum_{k=1}^n F_k(x) $$ $$ \begin{eqnarray} \mathbb{E}(Y^2) &=& \frac{1}{n^2} \sum_{k=1}^n \sum_{\ell=1}^n \mathbb{E}\left( I(X_k \leqslant x) I(X_\ell \leqslant x) \right) \\ &=& \frac{1}{n^2} \sum_{k=1}^n \sum_{\ell=1}^n \mathbb{P}(X_k \leqslant x, X_\ell \leqslant x) \end{eqnarray} $$ Therefore: $$\begin{eqnarray} \mathbb{Var}(Y) &=& \frac{1}{n^2} \sum_{k=1}^n \sum_{\ell=1}^n \left( \mathbb{P}(X_k \leqslant x, X_\ell \leqslant x) - F_{X_k}(x) F_{X_\ell}(x) \right) \\ &=& \frac{1}{n^2} \sum_{k=1}^n F_{X_k}(x) \left(1-F_{X_k}(x)\right) + \frac{2}{n^2} \sum_{1 \leqslant k < \ell \leqslant n} \left( F_{X_k,X_\ell}(x,x) - F_{X_k}(x) F_{X_\ell}(x) \right) \end{eqnarray}$$

For the case of independent variables in the sample, we get $$ \mathbb{Var}(Y_\text{indep}) = \frac{1}{n^2} \sum_{k=1}^n F_{X_k}(x) \left(1-F_{X_k}(x)\right) $$ For the case of identically distributed: $$ \mathbb{Var}(Y_\text{i.i.d.}) = \frac{1}{n} F_X(x) (1-F_X(x)) $$

share|improve this answer
    
Thank you. How about the case of $m$-independence where $X_i$ and $X_j$ are independent once $m$ apart. –  user12847 Apr 30 '12 at 6:30
    
@user12847 In that case, the formula for $\mathbb{Var}(Y)$ simplifies, in that $F_{X_k,X_\ell}(x,x) = F_{X_k}(x) F_{X_\ell}(x)$ for $|k-\ell| \geqslant m$. –  Sasha Apr 30 '12 at 12:02
    
Then the variance increases with positive correlation among random variables. Why? –  user12847 Apr 30 '12 at 19:57
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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