Suppose we have $2$ independent population parameters $p_1$ and $p_2$, such that the $90$% ( symmetric ) confidence intervals for $p_1$ and $p_2$ are given by $ (0.411, 0.498) $ and $(0.473, 0.567)$ respectively. Now, according to the original question ( copied verbatim ) , supposing that the $90 $% interval for $p_1$ lies entirely below the interval for $p_2$, how confident are we to conclude that $p_1 < p_2$?

Now, assuming that I have understood the problem correctly, I feel that this question does not make sense to me because $(0.411, 0.498)$ does not lie completely behind $(0.473,0.567)$. The question is, am I right in claiming so?

Next, assuming the problem is simply "How confident are we to conclude that $p_1 < p_2 $ ", given their $2$ $90$% symmetric confidence intervals as above, how would we go about approaching it, if it is even possible?


When we say that an interval has confidence 95% we mean that in repeated sampling we would find that intervals constructed in the same way would cover the true value of the parameter about 95% of the times: the confidence is the probability of coverage in repeated experimentation.

Now, I do not think you can talk about "How confident are we to conclude that $p_1<p_2$" when no confidence interval is involved: $p_1 < p_2$ that will be either true or false. Perhaps from a Bayesian point of view you could talk about the posterior probability that $p_1 < p_2$ given that something happens, but that is an entirely different sort of statement than what goes under the name of confidence.

  • 1
    $\begingroup$ Yes, so what? My answer was that you cannot properly talk of "the confidence that $p_1 < p_2$". $\endgroup$ – F. Tusell Apr 19 '20 at 16:20
  • $\begingroup$ ok, I understand your point now. Thank you! $\endgroup$ – ONG SEE HAI Apr 19 '20 at 16:30

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.