3
$\begingroup$

I have several discrete datasets with a small range (-2 to +2) which I want to test against the null hypothesis "the distribution is centred around zero".

The Wilcoxon signed-rank test was suggested to me as an appropriate choice, and some answers on this site back this up. The issue is that, for some datasets, a large portion of the pair deltas are 0, meaning most of the samples are discarded. The result is that some datasets pass the test at 95% confidence while others with nearly identical distributions fail it.

What would be a good alternative test for such a scenario?

$\endgroup$

1 Answer 1

1
$\begingroup$

This article:

On Neutral Responses (Zeros) in the Sign Test and Ties in the Wilcoxon–Mann–Whitney Test Ronald H Randles The American Statistician Vol. 55, Iss. 2, 2001

Would seem to apply directly to your question.

$\endgroup$
1
  • $\begingroup$ Thanks. For the record, the majority preference hypothesis seems the appropriate choice as suggested by the paper. $\endgroup$ Commented Sep 23, 2013 at 22:47

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.