gianni lavaredo
2012-Mar-26 19:32 UTC
[R] Different result with "kruskal.test" and post-hoc analysis with Nemenyi-Damico-Wolfe-Dunn test implemented in the help page for oneway_test in the coin package that uses multcomp
Dear Researchers,
Sorry for this email but I am not a statistician, and for this I have this
problem to understand. Thanks in Advance for help and suggestions.
Gianni
I have 21 classes (00, 01, 02, 04, ....,020) with different length. I did a
kruskal wall test in R with the following code
kruskal.test(m.class.l, m.class.length.lf)
Kruskal-Wallis rank sum test
data: m.class.l and m.class.length.lf
Kruskal-Wallis chi-squared = 34.2904, df = 20, p-value = 0.02423
the p-value result to be < of 0.05, and i read in different books and forum
that when the medium are significative difference to apply a post-hoc test
to check which classes are different.
i find this code in a R help:
http://stackoverflow.com/questions/2478272/kruskal-wallis-test-with-details-on-pairwise-comparisons
class <- m.class.length.lf
var <- m.class.l
dft <- data.frame(class,var)
NDWD <- oneway_test(var ~ class, data = dft,
ytrafo = function(data) trafo(data, numeric_trafo = rank),
xtrafo = function(data) trafo(data, factor_trafo = function(x)
model.matrix(~x - 1) %*% t(contrMat(table(x), "Tukey"))),
teststat = "max", distribution = approximate(B=1000))
### global p-value
print(pvalue(NDWD))
[1] 0.074
99 percent confidence interval:
0.05425181 0.09791886
### sites (I = II) != (III = IV) at alpha = 0.01 (page 244)
print(pvalue(NDWD, method = "single-step"))
when I apply the code for my data for see details for each pair
comparison, no class comparision is significative
> prova <- pvalue(NDWD, method = "single-step")
> prova.df <- data.frame(prova)
> prova.df$sing <- Signif(prova.df$V1)
> prova.df
V1 sing
01 - 00 1.000 ns
010 - 00 1.000 ns
011 - 00 1.000 ns
012 - 00 1.000 ns
013 - 00 1.000 ns
014 - 00 1.000 ns
015 - 00 1.000 ns
016 - 00 1.000 ns
017 - 00 0.996 ns
018 - 00 0.870 ns
019 - 00 1.000 ns
02 - 00 1.000 ns
020 - 00 0.642 ns
03 - 00 1.000 ns
04 - 00 1.000 ns
05 - 00 1.000 ns
06 - 00 1.000 ns
07 - 00 1.000 ns
08 - 00 1.000 ns
09 - 00 1.000 ns
010 - 01 1.000 ns
011 - 01 1.000 ns
012 - 01 1.000 ns
013 - 01 1.000 ns
014 - 01 1.000 ns
015 - 01 1.000 ns
016 - 01 1.000 ns
017 - 01 0.981 ns
018 - 01 0.737 ns
019 - 01 1.000 ns
02 - 01 1.000 ns
020 - 01 0.441 ns
03 - 01 1.000 ns
04 - 01 1.000 ns
05 - 01 1.000 ns
06 - 01 1.000 ns
07 - 01 1.000 ns
08 - 01 1.000 ns
09 - 01 1.000 ns
011 - 010 1.000 ns
012 - 010 1.000 ns
013 - 010 1.000 ns
014 - 010 1.000 ns
015 - 010 1.000 ns
016 - 010 1.000 ns
017 - 010 0.990 ns
018 - 010 0.774 ns
019 - 010 1.000 ns
02 - 010 1.000 ns
020 - 010 0.490 ns
03 - 010 1.000 ns
04 - 010 1.000 ns
05 - 010 1.000 ns
06 - 010 1.000 ns
07 - 010 1.000 ns
08 - 010 1.000 ns
09 - 010 1.000 ns
012 - 011 0.963 ns
013 - 011 1.000 ns
014 - 011 1.000 ns
015 - 011 1.000 ns
016 - 011 0.993 ns
017 - 011 0.673 ns
018 - 011 0.226 ns
019 - 011 0.869 ns
02 - 011 1.000 ns
020 - 011 0.074 ns
03 - 011 1.000 ns
04 - 011 1.000 ns
05 - 011 1.000 ns
06 - 011 1.000 ns
07 - 011 1.000 ns
08 - 011 1.000 ns
09 - 011 1.000 ns
013 - 012 1.000 ns
014 - 012 1.000 ns
015 - 012 1.000 ns
016 - 012 1.000 ns
017 - 012 1.000 ns
018 - 012 1.000 ns
019 - 012 1.000 ns
02 - 012 1.000 ns
020 - 012 1.000 ns
03 - 012 1.000 ns
04 - 012 0.991 ns
05 - 012 1.000 ns
06 - 012 1.000 ns
07 - 012 1.000 ns
08 - 012 0.995 ns
09 - 012 1.000 ns
014 - 013 1.000 ns
015 - 013 1.000 ns
016 - 013 1.000 ns
017 - 013 0.976 ns
018 - 013 0.707 ns
019 - 013 1.000 ns
02 - 013 1.000 ns
020 - 013 0.395 ns
03 - 013 1.000 ns
04 - 013 1.000 ns
05 - 013 1.000 ns
06 - 013 1.000 ns
07 - 013 1.000 ns
08 - 013 1.000 ns
09 - 013 1.000 ns
015 - 014 1.000 ns
016 - 014 1.000 ns
017 - 014 0.995 ns
018 - 014 0.839 ns
019 - 014 1.000 ns
02 - 014 1.000 ns
020 - 014 0.560 ns
03 - 014 1.000 ns
04 - 014 1.000 ns
05 - 014 1.000 ns
06 - 014 1.000 ns
07 - 014 1.000 ns
08 - 014 1.000 ns
09 - 014 1.000 ns
016 - 015 1.000 ns
017 - 015 0.978 ns
018 - 015 0.704 ns
019 - 015 1.000 ns
02 - 015 1.000 ns
020 - 015 0.383 ns
03 - 015 1.000 ns
04 - 015 1.000 ns
05 - 015 1.000 ns
06 - 015 1.000 ns
07 - 015 1.000 ns
08 - 015 1.000 ns
09 - 015 1.000 ns
017 - 016 1.000 ns
018 - 016 1.000 ns
019 - 016 1.000 ns
02 - 016 1.000 ns
020 - 016 0.992 ns
03 - 016 1.000 ns
04 - 016 1.000 ns
05 - 016 1.000 ns
06 - 016 1.000 ns
07 - 016 1.000 ns
08 - 016 1.000 ns
09 - 016 1.000 ns
018 - 017 1.000 ns
019 - 017 1.000 ns
02 - 017 1.000 ns
020 - 017 1.000 ns
03 - 017 0.916 ns
04 - 017 0.785 ns
05 - 017 0.999 ns
06 - 017 0.994 ns
07 - 017 1.000 ns
08 - 017 0.852 ns
09 - 017 0.994 ns
019 - 018 1.000 ns
02 - 018 0.905 ns
020 - 018 1.000 ns
03 - 018 0.536 ns
04 - 018 0.329 ns
05 - 018 0.883 ns
06 - 018 0.821 ns
07 - 018 0.914 ns
08 - 018 0.442 ns
09 - 018 0.798 ns
02 - 019 1.000 ns
020 - 019 0.981 ns
03 - 019 0.993 ns
04 - 019 0.952 ns
05 - 019 1.000 ns
06 - 019 1.000 ns
07 - 019 1.000 ns
08 - 019 0.978 ns
09 - 019 1.000 ns
020 - 02 0.677 ns
03 - 02 1.000 ns
04 - 02 1.000 ns
05 - 02 1.000 ns
06 - 02 1.000 ns
07 - 02 1.000 ns
08 - 02 1.000 ns
09 - 02 1.000 ns
03 - 020 0.229 ns
04 - 020 0.118 ns
05 - 020 0.647 ns
06 - 020 0.544 ns
07 - 020 0.678 ns
08 - 020 0.166 ns
09 - 020 0.496 ns
04 - 03 1.000 ns
05 - 03 1.000 ns
06 - 03 1.000 ns
07 - 03 1.000 ns
08 - 03 1.000 ns
09 - 03 1.000 ns
05 - 04 1.000 ns
06 - 04 1.000 ns
07 - 04 1.000 ns
08 - 04 1.000 ns
09 - 04 1.000 ns
06 - 05 1.000 ns
07 - 05 1.000 ns
08 - 05 1.000 ns
09 - 05 1.000 ns
07 - 06 1.000 ns
08 - 06 1.000 ns
09 - 06 1.000 ns
08 - 07 1.000 ns
09 - 07 1.000 ns
09 - 08 1.000 ns
[[alternative HTML version deleted]]
Possibly Parallel Threads
- Unexpected results using the oneway_test in the coin package
- Eliminating 'Unprintable ASCII' characters
- Asterisk 1.8.28-cert3, 1.8.32.1, 11.6-cert8, 11.14.1, 12.7.1, 13.0.1 Now Available (Security Release)
- Asterisk 1.8.28-cert3, 1.8.32.1, 11.6-cert8, 11.14.1, 12.7.1, 13.0.1 Now Available (Security Release)
- Nemenyi test as a post-hoc test to Kruskal Wallis
Wisdom of the Ancients
