1

I have a situation where I have very wide confidence intervals, and low rejection rates for a t test. I think that in general wider confidence intervals would be more likely to NOT reject the null hypothesis for a single sample t test since the null value is likely to be contained in the interval - since its so wide!

Is this result expected? Is there a name for whats going on here statistically?

kjetil b halvorsen
  • 77,844
user119563
  • 81
  • 1
  • 1
  • 3
  • Can you provide more context about the data? – Jon Dec 07 '16 at 01:07
  • Yes! This is from a simulation study where I am putting in very hevy tailed simulated t d-stributions (degrees of freedom 1, 2,3...) with sample sizes of 25 and 100 each. I am trying to see how the t test behaves when its presented with low degrees of freedom – user119563 Dec 07 '16 at 01:21
  • I can't say off the top of my head, but I would look through the paper mentioned in this post http://stats.stackexchange.com/questions/71953/relative-efficiency-of-wilcoxon-signed-rank-in-small-samples – Jon Dec 07 '16 at 01:34
  • If I understand you correctly you are using the t distribution as a population distribution and are applying a t test to the sample mean.. This can be confusing. But if I am right the a t with one or two degrees of freedom may be so heavy-tailed that the population mean doesn't exist as is also the case for the Cauchy distribution. In that case what is the parameter that you are trying to test for? How do you construct these confidence intervals? – Michael R. Chernick Dec 07 '16 at 08:07

0 Answers0