I am reading about confidence intervals and got stuck with this example from L. Wasserman's book titled "All of Statistics". Could anybody explain why PQ(θ ∈ C) = 3/4 in this example? Below is the paragraph from the book:

Let θ be a fixed, known real number and let X1, X2 be independent random variables such that P(Xi = 1) = P(Xi = -1) = 1/2. Now define Yi = θ + Xi and suppose that you only observe Y1 and Y2. Define the following interval that actually contains only one point:

enter image description here

You can check that, no matter what θ is, we have Pθ(θ ∈ C) = 3/4 so this is a 75 percent confidence interval. Suppose we now do the experiment and we get Y1 = 15 and Y2 = 17. Then our 75 percent confidence interval is {16}. However, we are certain that θ = 16. If you wanted to make a probability statement about θ you would probably say that P(θ ∈ C|Y1, Y2) = 1. There is nothing wrong with saying that {16} is a 75 percent confidence interval. But is it not a probability statement about θ.

  • $\begingroup$ I didn't fing the definition of $\theta$, is it the same as Q ? $\endgroup$
    – Pohoua
    Mar 31, 2020 at 11:56
  • $\begingroup$ Yes, it's the same. Sorry for that. I'll change. Do you have any ideas? $\endgroup$
    – kirgol
    Mar 31, 2020 at 12:18
  • $\begingroup$ There are only four possible outcomes: make a four-row table of the possibilities, work out their probabilities, and check which rows correspond to the event ${\Pr}_\theta(\theta\in C).$ $\endgroup$
    – whuber
    Mar 31, 2020 at 13:43
  • $\begingroup$ @whuber sorry, can you elaborate on that?. I only see that X1 and X2 can be 1 or -1 and in both cases they have P = 1/2. They are independent, thus P(X,Y) for all 4 possibilities is 1/4. How then we get 3/4 for Q? $\endgroup$
    – kirgol
    Mar 31, 2020 at 16:23
  • $\begingroup$ Because the event comprises three of the rows, whence its chance is three times 1/4. $\endgroup$
    – whuber
    Mar 31, 2020 at 16:25

1 Answer 1


You can dissociate cases :

  • if $X_1 \neq X_2$, which happens with probability $\frac{1}{2}$, then $X_1 = -X_2$ (since $X$ can only be $1$ or $-1$) and $Y_1 \neq Y_2$. So $C = \{\frac{Y_1 + Y_2}{2}\} = \{\theta\}$. So $\theta \in C$.

  • if $X_1 = X_2 = 1$, which happens with probability $\frac{1}{4}$, then $Y_1 = Y_2 = \theta + 1$ and $C = \{\theta\}$. So $\theta \in C$.

  • if $X_1 = X_2 = -1$, which happens with probability $\frac{1}{4}$, then $Y_1 = Y_2 = \theta - 1$, and $C = \{\theta - 2\}$. So $\theta \notin C$.

So the only possible case in which $\theta \notin C$ is when $X_1 = X_2 = -1$, which happens with probability $\frac{1}{4}$, so $P(\theta \in C) = 1 - \frac{1}{4} = \frac{3}{4}$

I think the point of Wasserman here is that the randomness lies in $C$ and not $\theta$. And indeed in the different cases considered, each time it is the confidence intervall $C$ which changes, not $\theta$.

  • $\begingroup$ Thank you very much for your explanation. I understood everything $\endgroup$
    – kirgol
    Apr 4, 2020 at 11:56

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.