[Rd] kruskalmc(pgirmess) (PR#10664)

sommerfeld at mpil-ploen.mpg.de sommerfeld at mpil-ploen.mpg.de
Wed Jan 30 11:20:22 CET 2008


Full_Name: Ralf Sommerfeld
Version: 2.3.1
OS: Windows XP
Submission from: (NULL) (193.175.214.49)


The kruskalmc function runs a multiple comparison test after a Kruskal-Wallis
test. I used the function for treatment comparison (default) as compared to a
comparison against a control. The alignment of names and obs.dif is however in
the wrong order!
++++++++++++++
The correct ordering (lines 13-19) is as follows:

         for (i in 2:(length(name))) {
             for (j in 1:(i-1)) {
                 vname <- c(vname, paste(name[j], "-", name[i], 
                   sep = ""))
		 indices <- rbind(indices, c( j, i))
             }
         }

instead of:

        for (i in 1:(length(name) - 1)) {
            for (j in (i + 1):length(name)) {
                vname <- c(vname, paste(name[i], "-", name[j], 
                  sep = ""))
                indices <- rbind(indices, c(i, j))
            }
        }
+++++++++++++++

I did not check the algorithm for ordering a two-sided or one-sided test against
a control.

All the best,
Ralf



More information about the R-devel mailing list