R-alpha: planned update of ctest

Peter Dalgaard BSA p.dalgaard@kubism.ku.dk
20 May 1997 20:18:02 +0200

Kurt Hornik <Kurt.Hornik@ci.tuwien.ac.at> writes:

> I am contemplating improving my ctest package as follows:
> * Add exact p,q,r,s functions for the Wilcoxon distribution, and change
> the test accordingly (make `exact' work).
> * Make Fisher's test work for tables larger than 2 by 2.
> * Perhaps add an exact unconditional test for 2 by 2 tables?
> * Perhaps add something on estimating/testing relative risk and odds?
> As clearly I'd like to minimize my efforts I'd be happy about any
> suggestion for availability of code, and for a good interface for the
> last idea.

As usual, all you'll get is suggestions for more enhancements:

* Order-based statistics with scores different from ranks (i.e. Van
der Waerdens test, etc.)

* Logrank test (a bit hefty to pull out the survival package for that)

* Stratified tests. I actually need those (Wilcoxon and K-W versions)
for a client, so I might put a new student computational assistant on
it. Not sure whether that will break her in or just break her,

Re. relative risks, you can quite easily setup an exact test for 
RR=r (I forget, but it's something like multiplying the point
probabilities in Fishers test with r^x). Consequently, you can also
compute confidence intervals. One snag is the usual mess with the
definition of the two-sided p-value. I have previously used two sided
intervals consisting of those values that weren't significantly too
big [1] at the alpha/2 level nor significantly too small at the same
level, but I'm not altogether sure that holds.

[1] Yeah, yeah. "...for which the data were not significantly too
small, using a one-sided test", etc.
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch