adjustments for wilcoxon signed rank test

[neptunhiker]

Hi,

I would like to replicate the results of the wilcoxon signed rank tests for equal medians by hand to confirm that I understand the test procedure. Unfortunately I get slightly different results which is probably due to the fact that my calculations don't include adjustments for ties and continuity. I have read in the documentation that the adjustments implemented in eviews follow Handbook of Sheskin (1997) and Conover (1980). Unfortunately I cannot get a hold of the two sources. Could someone therefore please poste the adjustments here?

Thank you very much.

Kind regards

[EViews Glenn]

We don't normally do this, but...

For continuity, take absolute value of the numerator and subtract .5.
Let t_i be the number of ties for the i-th rank. Then in the denominator of your expression subtract sum_i (t_i^3 - t_i) / 48 prior to taking the square root.

[neptunhiker]

Hi Glenn,

Thanks a lot. That really helped me out. I do now get the same results after adjustments.