I understand that pivotal quantities are usually found for finding confidence intervals via inversion. However, I am wondering if there is an intuition behind a pivotal quantity having to depend on our data and parameters $(X, \theta)$ while having a distribution that is free of the parameters. Is it so that we can isolate the parameter $\theta$ by itself in the confidence interval? If so, what property of the pivotal quantity allows this?


1 Answer 1


Yes. Essentially, the use of pivotality is that we obtain a statement of the sort $$1-\alpha=P\{\alpha/2\text{%-quantile}\leq\text{pivotal quantity}(X,\theta)\leq(1-\alpha/2)\text{%-quantile}\}$$ which holds for any $\theta$ so that we do not need to attach a $\theta$ to $P$ (as in $P_\theta$). We may then, as you say, rearrange the statements within the curly braces until we have isolated $\theta$. Since the equality still says $$1-\alpha=P\{...\}$$ we have derived a statement which holds for any $\theta$.

  • 1
    $\begingroup$ I just want to note that it is also useful to have one that removes dependence on nuisance parameters such as the variance in in the the case of the t test for the mean of a normal distribution. Also it is not always possible to construct a pivotal quantity. $\endgroup$ Dec 31, 2016 at 17:00

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.