Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The confidence interval for the mean of a random variable $Y$ has coverage $1-\alpha$ which I am trying to show. Starting from $$\widehat{E(Y)} - q_{1-\frac{\alpha}{2}}\sqrt{\frac{\widehat{Var(Y)}}{n}} \leq E(Y) \leq \widehat{E(Y)} + q_{1-\frac{\alpha}{2}}\sqrt{\frac{\widehat{Var(Y)}}{n}} $$ I re-arranged this to $$- q_{1-\frac{\alpha}{2}} \leq \sqrt{n} \left( \frac{\widehat{E(Y)}-E(Y)}{\sqrt{Var(Y)}} \right) \leq q_{1-\frac{\alpha}{2}} $$ Then by the central limit theorem $\sqrt{n} (\widehat{E(Y)} - E(Y)) \rightarrow N(0,Var(Y))$. Then I use the continuous mapping theorem. Given that $\widehat{Var(Y)} = \frac{1}{n}\sum^{n}_{i=1} Y_i^2 - (\frac{1}{n}\sum^n_i Y_i)^2$, define the function $m:(x,y) \rightarrow x - y^2$. Then $$x \equiv \sum^{n}_{i=1} Y_i^2 \rightarrow E(Y^2)$$ and $$y \equiv \sum^{n}_{i=1} Y_i \rightarrow E(Y)$$ by the weak law of large numbers. Then $\widehat{Var(Y)} = g(x,y) \rightarrow g(E(Y^2),E(Y)) = Var(Y)$ is a consistent estimator of the variance. Hence $$\sqrt{n} \left( \frac{\widehat{E(Y)}-E(Y)}{\sqrt{Var(Y)}} \right) \rightarrow N(0,1)$$

But now I am stuck in how to proceed to obtain the final coverage rate. If someone could give me a hint on how to get there, this would be greatly appreciated.


Thanks for pointing out the typo. I think I understand the hint. This should be $$ \begin{align} Pr \left(\sqrt{n} \left( \frac{\widehat{E(Y)}-E(Y)}{\sqrt{\widehat{Var(Y)}}} \in (-q_{1-\frac{\alpha}{2}}, q_{1-\frac{\alpha}{2}} \right) \right) &= Pr(N(0,1) \in (-q_{1-\frac{\alpha}{2}}, q_{1-\frac{\alpha}{2}} )) \newline &= Pr(N(0,1) \leq q_{1-\frac{\alpha}{2}}) - Pr(N(0,1) > q_{\frac{\alpha}{2}}) \newline &= \Phi (q_{1- \frac{\alpha}{2}}) - \Phi (q_{\frac{\alpha}{2}}) \newline &= 1 - \frac{\alpha}{2} - \frac{\alpha}{2} \newline &= 1 - \alpha \end{align} $$

share|improve this question
up vote 6 down vote accepted

Probably because you got the quantiles wrong (I suppose your $q(\cdot )$ are the quantiles of the standard normal), which should be $q_{1-\frac{\alpha}{2}}$. Basically what you want to show is that $$Pr\left( E(Y) \in \left( \widehat{E(Y)} - q_{1-\frac{\alpha}{2}} \sqrt{\frac{\widehat{Var(Y)}}{n}}, \widehat{E(Y)} + q_{1-\frac{\alpha}{2}} \sqrt{\frac{\widehat{Var(Y)}}{n}} \right) \right) $$ which is equal to showing that $$Pr\left( \sqrt{n}\left( \frac{\widehat{E(Y)}-E(Y)}{\sqrt{\widehat{Var(Y)}}} \right) \in \left( -q_{1-\frac{\alpha}{2}}, q_{1-\frac{\alpha}{2}} \right) \right) $$ You already know that $\sqrt{n}\left( \frac{\widehat{E(Y)}-E(Y)}{\sqrt{\widehat{Var(Y)}}} \right)$ is distributed as $N(0,1)$ so this should be straight forward. A trick that will help is to know that $-q_{1-\frac{\alpha}{2}}=q_{\frac{\alpha}{2}}$ in order to figure out the probabilities that $N(0,1) \leq q_{1-\frac{\alpha}{2}}$ and $N(0,1) > q_\frac{\alpha}{2}$

share|improve this answer

Your Answer


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.