Possibly remove pupsilon from sr_unpaired_test

[shabbychef]

The pupsilon code is suspect, as the higher order cumulants may not be small enough to support truncation. Possibly this code should be replaced by a simulation based approach?

[JoeOD]

Hello, I see there are some open issues for the sr_unpaired_test function. Would you consider it safe/advisable to use the function?

[shabbychef]

I would guess that this test currently does not reach the nominal type I rate, but I have not had the time to perform the simulations. Perhaps I can do those easily.

[shabbychef]

There may be problems with smaller sample sizes or in
the case of multiple contrasts or multiple SRs, but
this looks good to me so far:

library(SharpeR)
set.seed(1234)
allpv <- replicate(1000,
{
  x1 <- rnorm(100)
  x2 <- rnorm(50)
  srs <- list(as.sr(x1),as.sr(x2))
  sr_unpaired_test(srs)$p.value
})
plot(ecdf(allpv))
# or:
library(txtplot)
xv <- ppoints(1000)
yv <- ecdf(allpv)(xv)
txtplot(xv,yv,width=80)

    +--+-------------+-------------+------------+-------------+------------+---+
  1 +                                                                   ****   +
    |                                                                ****      |
    |                                                           ******         |
0.8 +                                                       *****              +
    |                                                   *****                  |
    |                                               *****                      |
0.6 +                                           *****                          +
    |                                       *****                              |
    |                                   *****                                  |
    |                               *****                                      |
0.4 +                           *****                                          +
    |                       *****                                              |
    |                   *****                                                  |
0.2 +              ******                                                      +
    |          *****                                                           |
    |      *****                                                               |
  0 +  *****                                                                   +
    +--+-------------+-------------+------------+-------------+------------+---+
       0            0.2           0.4          0.6           0.8           1