Hi all,
I am a biostatistician and I have developed my own
ranking system for clinical data. I would like to test
the efficiency of it w.r.t. to other ranking systems.
I would like to simulate the data and after assigning
ranks to my observed scores(after neglecting
dropouts), observe the type I error. If I want to do a
Kruuskal Wallis type of test, what test statistic
should I use to test for a difference of "delta"
between the two groups (I know "delta" since the data
is generated with that difference). The default K-W
test statistics test for a NULL difference and I want
to see how frequently my ranking system rejects the
correct null hypothesis of a "delta" difference.
(So my data is now in 2 arrays of unequal lengths due
to dropouts, which have the ranks for drug and
placebo.)
Thank you in advance for your help.
~S
