[R] anomalous result for wilcox.exact in exactRankTests

[Peter Dalgaard]

Roger Levy <rog at stanford.edu> writes:

> Hi,
> 
> In the exactRankTest package, I've become aware that you can get
> anomalous p-values (i.e., above 1) from the wilcox.exact method, as in:
> 
>   > wilcox.exact(c(-0.6,0.8,-0.5))
> 
>         Exact Wilcoxon signed rank test
> 
>   data:  c(-0.6, 0.8, -0.5) 
>   V = 3, p-value = 1.25
>   alternative hypothesis: true mu is not equal to 0 
> 
> This is disturbing.  Has anyone encountered this before, and if so is
> there an obvious reason why this should happen?

Presumably by counting the observed value towards both tails:

> dsignrank(0:6,3)
[1] 0.125 0.125 0.125 0.250 0.125 0.125 0.125
> sum(dsignrank(0:3,3))+sum(dsignrank(3:6,3))
[1] 1.25

which looks a bit odd, but of course only happens when the correct
p-value is 1.0, so it is fairly harmless.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907