R help - May 2013 - T statistic exactRankTests package

I understand this might be a silly question, but I have problems
understanding the statistic term returned in most permutation packages.
for example:

x<-c(8,4,6)
y<-c(5,3,4)
 res<-exactRankTests::perm.test(x,y,paired=TRUE)
         1-sample Permutation Test
data:  x and y
T = 6, p-value = 0.25
alternative hypothesis: true mu is not equal to 0

attributes(res)
$names
[1] "statistic"   "p.value"     "pointprob"  
"null.value"  "alternative"
"method"      "data.name"
$class
[1] "htest"

res$statistic
T
6

I cannot figure out what the T stands for. I understand it is "T
statistic"
but it is not the t-score.
Can some one tell me what formula is used for the T statistic.
In general is there some way to look for what function is used to calculate
the statistics?

Thanks

	[[alternative HTML version deleted]]

On 2013-05-30 04:20, Al Zou wrote:
> I understand this might be a silly question, but I have problems
> understanding the statistic term returned in most permutation packages.
> for example:
>
> x<-c(8,4,6)
> y<-c(5,3,4)
>   res<-exactRankTests::perm.test(x,y,paired=TRUE)
>           1-sample Permutation Test
> data:  x and y
> T = 6, p-value = 0.25
> alternative hypothesis: true mu is not equal to 0
>
> attributes(res)
> $names
> [1] "statistic"   "p.value"     "pointprob"  
"null.value"  "alternative"
> "method"      "data.name"
> $class
> [1] "htest"
>
> res$statistic
> T
> 6
>
> I cannot figure out what the T stands for. I understand it is "T
statistic"
> but it is not the t-score.
> Can some one tell me what formula is used for the T statistic.
The formula is simple: T is the sum of the positive paired differences.
You could check Wikipedia for "permutation test". There must be plenty
of online info but I'm too lazy to check.
> In general is there some way to look for what function is used to calculate
> the statistics?
>
Just look at the code of perm.test.default (it's not complicated).
Type

   perm.test.default

to see the code or get the package source and study the code.


Peter Ehlers