0

@FrankHarrell has suggested that “A good nonparametric approach to getting confidence intervals for means and differences in means is the bootstrap.” That said, my question is how to interpret dotplots of bootstrap means with 95% CIs. Cumming and Finch suggested in their paper Inference by Eye that for 2 independent samples there can be a 0.58 proportion overlap yielding p = 0.047 as you can see in their Figure 4enter image description here.

Do the Cumming/Finch recommendations on how much overlap is acceptable apply to bootstrap means?

FTF
  • 71
  • 5
  • For a difference of means wouldn't you compute the observed difference and then the differences for the bootstrap samples and form confidence intervals using these values for the difference itself, rather than for individual means and trying to interpret overlap? – Gavin Simpson Apr 09 '16 at 16:45
  • My question is not about computing the differences, but about interpreting a dotplot of 2 bootstrap means with 95% CI. – FTF Apr 09 '16 at 17:20
  • No, I understand that; my point was, why even bother with trying to interpret overlaps when you are doing the bootstrap anyway? I haven't fully read the paper you linked to but it seems like they are suggesting ways to calibrate reading things off plots other people have produced. If you can avoid this by computing a CI for the thing of interest, that surely has to better? As to the original Q, I doubt these things hold for bootstrap samples, unless the bootstrap distribution is well approximated by a Gaussian distribution. – Gavin Simpson Apr 09 '16 at 17:37
  • My question is not directed to the person who has access to the data and can do calculations, it is directed to somebody reading a paper that looks like the figure but was generated with bootstrapped means with 95%CI. How should he interpret it? Since @FrankHarrell posed the recommendation, I hope that he has some thoughts on the matter. – FTF Apr 09 '16 at 18:00
  • There's nothing special about bootstrapped CIs versus other procedures to compute CIs. As far as interpreting overlap among them, maybe the thread at http://stats.stackexchange.com/questions/18215 answers your question? – whuber Apr 09 '16 at 18:07

0 Answers0