Subject: | A bug in Statistics::Test::WilcoxonRankSum; |
Date: | Wed, 16 Feb 2011 16:39:08 -0800 (PST) |
To: | <bug-statistics-test-wilcoxonranksum [...] rt.cpan.org> |
From: | "Rong Chen" <rchen1 [...] stanford.edu> |
Dear Sir,
I am using Statistics::Test::WilcoxonRankSum and find that if I change all
values to 1-old values, and your test will change form significant to
non-significant.
The script is attached.
Would you advise?
Thank you.
Rong
############################################
Rong Chen, Ph.D.
Bioinformatics Scientist
Division of Systems Medicine
Dept. of Pediatrics
MSOB X155
251 Campus Drive
Stanford University
Stanford, CA 94305
Tel: 858-837-2265
Fax: 650-724-2259
http://buttelab.stanford.edu/doku.php?id=people:rong
#############################################
bmir-cluster /home/rchen1/project/test 131 : perl test.pl
dataset1: 0.465 0.453 0.486
dataset2: 0.891 0.888 0.838 0.783 0.553 0.766 0.5 0.493
----------------------------------------------------------------
dataset | n | rank sum: observed / expected
----------------------------------------------------------------
1 | 3 | 6 / 16
----------------------------------------------------------------
2 | 8 | 60 / 44
----------------------------------------------------------------
N (size of both datasets): 11
Probability: 0.012121, exact
Significant (at 0.05 level)
Ranks of dataset 1 are lower than expected
all values are subtrcated by 1
dataset1: 0.535 0.547 0.514
dataset2: 0.109 0.112 0.162 0.217 0.447 0.234 0.5 0.507
----------------------------------------------------------------
dataset | n | rank sum: observed / expected
----------------------------------------------------------------
1 | 3 | 30 / 16
----------------------------------------------------------------
2 | 8 | 36 / 44
----------------------------------------------------------------
N (size of both datasets): 11
Probability: 2.000000, exact
Not significant (at 0.05 level)
Message body is not shown because sender requested not to inline it.