I see I am too late to comment :)
But commenting after the fact, just wish to say that I like the changes.
Specially the mentioning of "exact" in the test name.
Floating point prevision is very nicely implemented too.
My only worry is that it will not serve new/lay users that may be in the
bi
> Martin Maechler
> on Thu, 12 Dec 2019 17:20:47 +0100 writes:
> Karolis Koncevičius
> on Mon, 9 Dec 2019 23:43:36 +0200 writes:
>> So I tried adding Infinity support for all cases. And it
>> is (as could be expected) more complicated than I
>> thought.
> Karolis Koncevičius
> on Mon, 9 Dec 2019 23:43:36 +0200 writes:
> So I tried adding Infinity support for all cases.
> And it is (as could be expected) more complicated than I thought.
"Of course !" Thank you, Karolis, in any case!
> It is easy to add Inf support for
So I tried adding Infinity support for all cases.
And it is (as could be expected) more complicated than I thought.
It is easy to add Inf support for the test. The problems start with
conf.int=TRUE.
Currently confidence intervals are computed via `uniroot()` and, in the
case of infinities, we
I'd like to ask the developers to include some exact computation for
ties into wilcox.test(). Just try
wilcox.test(c(1,1,5),c(10,11))
wilcox.test(c(1,2,5),c(10,11))
The p-values differ significantly.
But if I try
library(exactRankTests)
wilcox.exact(c(1,1,5),c(10,11))
wilcox.exact(c(1,2,5)
Thank you for a fast response. Nice to see this mailing list being so
alive.
Regarding Inf issue: I agree with your assessment that Inf should not be
removed. The code gave me an impression that Inf values were
intentionally removed (since is.finite() was used everywhere, except for
paired ca
> Karolis Koncevičius
> on Sat, 7 Dec 2019 20:55:36 +0200 writes:
> Hello,
> Writing to share some things I've found about wilcox.test() that seem a
> a bit inconsistent.
> 1. Inf values are not removed if paired=TRUE
> # returns different results (Inf is removed
Thank you for responding, and so quickly at that.
Yes, I do understand that this is a floating point issue.
However, since wilcox.test() works on ranks this might be a bit
dangerous in my opinion. Maybe more so than for magnitude based tests.
Any small precision error will be ranked and it becom
Your second issue seems like a more or less unavoidable floating-point
computation issue. The paired test operates by computing differences
between corresponding values of x and y.
It's not impossible to try to detect "almost-ties" (by testing for
differences less than, say, sqrt(.Machine$do
Hello,
Writing to share some things I've found about wilcox.test() that seem a
a bit inconsistent.
1. Inf values are not removed if paired=TRUE
# returns different results (Inf is removed):
wilcox.test(c(1,2,3,4), c(0,9,8,7))
wilcox.test(c(1,2,3,4), c(0,9,8,Inf))
# returns the same result (Inf
10 matches
Mail list logo
