Exact Distributions for Rank and Permutation Tests

Computes exact conditional p-values and quantiles using an implementation of the Shift-Algorithm by Streitberg & Roehmel.


0.8-29 (01.03.2017)

register C routines

0.8-28 (20.02.2015)

register all methods

0.8-27 (24.09.2013)

    p-values were > 1 (in fact, 1 of course) when the statistic was
    equal its expectation in the paired case.

0.8-26 (02.09.2013)

fix NAMESPACE issues

0.8-25 (21.05.2013)

    perm.test(x, y, paired = TRUE, exact = TRUE, tol = NULL) did not map the
    scores correctly into integers and gave wrong p-values;
    spotted by Lewis Halsey <[email protected]>

0.8-24 (11.01.2013)

startup message in .onAttach

0.8-23 (11.10.2012)

cpermdist2 R function for maxstat

0.8-22 (30.08.2011)

    spaces in startup message

0.8-21 (23.08.2011)

    GPL (>= 2) only

0.8-20 (02.05.2011)

qnorm(..., lower -> lower.tail)

0.8-19 (13.09.2010)

    fix dependencies

0.8-16 (29.08.2008)

    fix old style random seed assignments

0.8-16 (08.10.2007)

documentation update

0.8-15 (05.10.2006)

    clarify dependencies on stats

0.8-14 (04.10.2006)

    null.value had different names

0.8-13 (23.08.2006)

fire a message that `exactRankTests' is no longer under development
and `coin' should be used.

0.8-12 (02.02.2006)

    remove *.Rout from R/

0.8-11 (02.06.2005)

    `mu' was ignored for one-sample version of `perm.test' 
    (spotted by <[email protected]>)

0.8-10 (16.03.2005)

`delay' is deprecated in R-2.1.0

0.8-7 (27.09.2004)

use a grid search instead of optimize in order to determine
    fact in `findfact'

0.8-6 (16.12.2003)

do not import(ctest) in Namespace and load package `stats' instead
for R versions >= 1.9.0

for the moment, copy `ansari.c' from package `stats'.

0.8-5 (03.11.2003)

fix documentation bugs

    0.8-4 (18.09.2003)

    perm.test(x, ...) failed when missing values where supplied.

0.8-3 (03.09.2003)

use explicit RNGkind("Wichmann-Hill", "Kinderman-Ramage") 
in all regression tests.

0.8-2 (30.08.2003)

use canonial header files

0.8-2 (18.09.2003)

perm.test(x, ...) failed when missing values where supplied.

0.8-1 (24.04.2003)

make -Wall happy

0.8-0 (23.04.2003)

Monte-Carlo based computation of the null-distribution now possible

use `.Call' instead of `.C' for interfacing with C

Code cleanup, documentation cleanup

Move regression tests from example sections to ./tests/reg-tests.R

Conover-Salsburg scores added, Savage scores added

more examples:	neuropathy (pain scores)
		clodronate (randomized trial)
		ears (ventilating tubes)
		sal (serum antigen level)
		bloodp (blood pressure)
		lungcancer (well,...)

prepare for R-1.7.0: set RNGversion("1.6.2") for regression tests
		     put the package into a NAMESPACE

0.7-7 (20.01.2003)

Rdiff reported differences due to cosmetic changes

0.7-6 (19.01.2003)

ansari.exact added: Ansari-Bradley test with exact conditional
    more examples: 	ASAT (proof-of-safety or equivalence),
		ocarcinoma (exact conditional logrank test),
		rotarod (WMW-test benchmark from  AmStat 54(1) )

0.7-5 (13.12.2002)

CRAN pedantics

0.7-4 (13.12.2002)

check if scores contain negative values for paired samples

0.7-3 (22.11.2002)

PACKAGE argument to Call

0.7-2 (18.11.2002)

force irank to return integers

0.7-1 (17.09.2002)

improve documentation
bugfix: pperm for two.sided p-values (paired / !paired) was wrong
    	if q == expectation
limits on the scores removed: [dpq]perm only limited by
    	the machine memory
cscores.Surv optimized, example with censored data added to

0.7-0 (13.09.2002)

generic `cscores' for the computation of scores added
m=NULL allowed in equiscores for some configurations

0.6-12 (23.07.2002)

bugfix for paired samples: do not substract the minimum of the
    scores anymore.

0.6-11 (03.07.2002)

use .C( ..., PACKAGE="exactRankTests")

0.6-10 (23.05.2002)


0.6-9 (15.04.2002)

added bivariate rank test `br.test' for testing only

0.6-8 (15.04.2002)

bug in perm.test.default for real-valued scores fixed

0.6-7 (10.04.2002)

replace optim in equiscores by optimize as suggested by warning in

0.6-6 (06.03.2002)

checkTnF was still unhappy with stuff in ./tests

0.6-5 (06.03.2002)

make checkTnF happy

0.6-4 (20.02.2002)

conf.int for paired samples was broken in wilcox.exact

p.value for alternative == "less" and "greater" now add to 1 for
    asymptotic distribution

0.6-3 (13.02.2002)

STATISTIC in perm.test had a missing name attribute

0.6-2 (11.02.2002)

R CMD check insits on \n at the end of C-files

0.6-1 (06.02.2002)

length(x) must be smaller than length(y) in perm.test(x,y), give a
    warning and interchange if not

asymptotic p-values in perm.test( ..., paired=FALSE, ...) were
wrong, fixed

0.6-0 (10.12.2001)

release for R-1.4.0

0.5-9 (08.12.2001)

map real scores into 1:N for perm.test
perm.test and wilcox.exact are generic and have a formula interface	
fixed the most painful documentation bugs

0.5-7 (06.11.2001)

asymptotic p-values for perm.test and wilcox.test, coincides with

0.5-7 (06.11.2001)

new interface to pperm (computes now all kinds of p.values)
therefore wilcox.exact and perm.test changed. several minor code-
and documentation fixes.

0.5-6 (31.10.2001)

bugfix in one-sided conf.ints paired wilcox.exact
examples from Hollander & Wolfe added to wilcox.exact
check for conf.ints and estimators. 
R CMD check will only work with R-1.4.0-devel and modified 

0.5-5 (29.10.2001)

estimated corrected (PR#1150)

0.5-4 (27.10.2001)

pperm has `side' argument: compute P(T <= T) or P(T >= t)

0.5-3 (25.10.2001)

perm.test and globulin-example added

0.5-2 (27.09.2001)

codoc fixes for R-1.4.0

0.4-6 (26.06.2001)

fixed confints, added tests:
- all.equal(wilcox.exact, wilcox.test)
    - estimator \in confint
    DEPENDS: R(>=1.3.0)

0.4-5 (09.06.2001)

Depend broke update.packages -> fixed for R-1.2.3
!!! REMINDER: change version$minor == "3.0" if R-1.3.0 is out !!!
added paired Wilcoxon example from StatXact manual, coincides :-)

0.4-3 (07.06.2001)

moved ./inst/test*.R -> ./tests

0.4-2 (07.06.2001)

documentation fixed

0.4-1 (25.05.2001)

sample estimates as in wilcox.test added to wilcox.exact

0.4-0 (16.05.2001)

use eqn. (8.19) from the StatXact manual p. 165 for twosided p-value
if paired.

0.3-8 (03.05.2001)

c++ comments removed from permdist.c

0.3-7 (12.04.2001)

pperm2 now computes correct expectation for paired case.
Now confidence intervals for x - y are computed (as `wilcox.test'
in R-1.2.3 will do)	

0.3-6 (20.02.2001)

typos in cperm.Rd fixed

0.3-4 (06.01.2001)

concentrated man pages.

0.3-3 (20.12.2000)

added some checks for non-sense parameter configurations. This one
goes out to CRAN/contrib.

0.3-2 (18.12.2000)

added pperm2 which computes two-sided p-values the way StatXact
does. Added example from StatXact manual page 211 on diastolic blood
pressure, the results coincied.  

0.3-1 (17.12.2000)

renamed "exactDistr" to "exactRankTests" (this is easier to
pronounce). After a discussion with Kurt last Friday, this version 
will move to CRAN/contrib. 

0.2-4 (05.12.2000)

typos fixed, system.file()[1] in examples 

0.2-3 (04.12.2000)

uhh, 0.2-2 was buggy :-(
bugs (hopefully) fixed, more intelligent determination of the
factor to multiply real (!) valued scores with. It is now possible
to give a maximum error by "tol" on the quantile scale. Default is 
tol = 0.01 which should be ok for most situations. 

0.2-2 (29.11.2000)

rational scores now possible -> this results in an approximation
only. Can be influenced by "fact".

0.1-9 (25.10.2000)

bugfixes for pperm(paired): 0 is a possible realization

0.1-8 (24.10.2000)

added the paired sample situation, now ./inst/test1.R for the
comparison with [pq]signrank, ./inst/test2.R with  [pq]wilcox

0.1-7 (23.10.2000)

bug fixes (H as double), test.R in ./inst for testing with

0.1-6 (22.10.2000)

hacked cpermdist in C, now computations a lot faster

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.8-32 by Torsten Hothorn, 20 days ago

Browse source code at https://github.com/cran/exactRankTests

Authors: Torsten Hothorn [aut, cre] , Kurt Hornik [aut]

Documentation:   PDF Manual  

Task views: Survival Analysis

GPL (>= 2) license

Depends on stats, utils

Suggests survival

Imported by Coinprofile, maxstat.

Suggested by DiscreteQvalue, MKinfer, MKmisc.

See at CRAN