0
$\begingroup$

Most scientific papers present data using mean (95%CI lower bound, 95%CI upper bound) but I have not seen mean ± 95% CI (upper bound - mean). Can I use the latter to present my results?

e.g.,

mean = 1.0
CIlb = 0.8
CIub = 1.2

Can I present the result as 1 ± 0.2 (mean ± 95%CI)or must I present as 1.0 (0.8, 1.2)?

$\endgroup$
5
  • $\begingroup$ Can you give a concrete example? $\endgroup$ Commented Jul 6, 2023 at 14:20
  • $\begingroup$ yes, I just added one $\endgroup$
    – Elizabeth
    Commented Jul 6, 2023 at 14:20
  • 3
    $\begingroup$ You can do that... if the CI is symmetric. $\endgroup$ Commented Jul 6, 2023 at 14:21
  • $\begingroup$ @user2974951 How is the poor reader supposed to understand either expression, exactly? There appear to be multiple reasonable interpretations of them, and "mean ± 95%CI," which explicitly appears to sum a number with an interval about that number is either nonsensical or wrong. $\endgroup$
    – whuber
    Commented Jul 6, 2023 at 17:41
  • $\begingroup$ I suspect that credible intervals would serve much better for the kind of result summary that you have in mind. $\endgroup$
    – Durden
    Commented Jul 7, 2023 at 0:14

1 Answer 1

3
$\begingroup$

Not all CIs are symmetrical, so the tradition is to report a confidence interval as an interval ("from LL to UL", or "LL,UL")

In your example, the CI is symmetrical so the plus/minus nomenclature is not ambiguous and is easy to understand. But it is not standard. You describe it as "mean ± 95% CI" but you really mean "mean ± margin of error", which is a better way to describe your syntax.

$\endgroup$

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.