2
$\begingroup$

When reporting results of a bootstrapped independent samples t-test, can I just report the t statistic alongside the bootstrapped p value? Or do I also need to report confidence intervals?

$\endgroup$
  • 1
    $\begingroup$ It strongly depends on want do you want to say. If you want to report only the difference - the "t" and p-value may be sufficient. If you want to emphasize confidence - add CI. Either way I recommend you to read the "author guidelines" of journal where you want to publish your work. $\endgroup$ – Ladislav Naďo Mar 30 '14 at 21:26
1
$\begingroup$

You can do pretty much as you see fit to do.

Broadly speaking, it may be reasonable to either report or not report a confidence interval depending on what you want to achieve/convey.

If you really mean "what will some supervisor/colleague/journal editor/reviewer accept?", that's a very different question, which would depend on things we're not privy to (which supervisor or colleague? which journal? ... and what do their guidelines say?)

I'd certainly report both the statistic and the p-value, as you suggested (and probably something more conventional, such as an ordinary t-test p-value or a permutation test p-value, so people could see what difference it makes, if any).

$\endgroup$
  • $\begingroup$ Thank you!! For dissertation findings for now, but hoping to publish later on. $\endgroup$ – Sue Mar 31 '14 at 5:39
  • $\begingroup$ Expectations can vary a lot from one application area to the next, one journal to the next, one editor to the next and one referee to the next, often with little to no justification. (Much of the received wisdom isn't necessarily especially wise.) That makes it very hard to generalize. $\endgroup$ – Glen_b -Reinstate Monica Mar 31 '14 at 5:51

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.