0

Some kind of a weird question, but I couldn't find anything on this topic. In our statistic-course we were using the excel addin WinSTAT for all of our concerns and well yes it's easy to use and didn't had any problems so far.

But when I tried to compute for example for the "Wilcoxon signed-rank test" in R the results always differ from the WinSTAT results. I'm not sure if this is my mistake or from WinSTAT. I can't imagine that R has an error in such a basic functionality.

My call in R: wilcox.test(data$1960, data$2002, paired = T, exact = F)
The p-value = 0.0002876
In WinSTAT the p-value is: 0,000293052520203

Here is the data.

So the question is it my fault or WinSTAT's?

EDIT: One more info: when I compare the results for example for a "paired t-Test" everything is fine and without any problems.

basto
  • 35
  • 1
  • 10
  • Probably numerical rounding issues. These results are statistically identical in that both will strongly reject the NULL at any traditional level of confidence. See [this post](https://stackoverflow.com/questions/9508518/why-are-these-numbers-not-equal) for some additional background info. – lmo May 28 '18 at 20:20
  • There are several different ways to calculate p-values for discrete tests like that one: normal approximations, exact enumeration, etc. R's `wilcox.test` function has arguments `exact` and `correct` which modify its calculations; probably WinSTAT uses a different method than the R default. – user2554330 May 29 '18 at 01:04
  • @user2554330: tried all kinds of arguments already, but nothing really worked – basto May 29 '18 at 06:06
  • @Imo: thanks for the info, but I guess I can't change anything and have to use WinSTAT instead of R.. It's a pity! – basto May 29 '18 at 06:25

0 Answers0