Dear all,
I want to evaluate the effect of a treatment on lameness in horses.
Lameness was evaluated at 3 occasions (1: before; 2:after; 3:quite a time
after treatment) by a panel of 3 veterinary surgeons using an ordinal scale
(0,1,2,3,4,5).
10 horses participated in the study. Two legs (left & right frontleg) were
evaluated and treated in 8 animals and only 1 leg (left or right frontleg)
in 2 animals.
My first step was to add the lameness-scores of the panel for each
horse/leg/occasion to a single value ("overall lamenes score", ordinal,
range 0-15) since I am not interested in differences between the
veterinarians.
Then I want to test wether there are diffences in lameness on the three
occasions. There is no gaussian-distribution of lameness data so my first
try was the non-parametric Wilcoxon test but there are (at least) two
problems:
#1) samples are not independent since 8 horses contribute 2 legs (= 2 rows).
#2) since there are 3 occasions I would have to perform wilcoxon-tests on
all 3 possible pairs (1-2; 1-3; 2-3) which my stat-docu tells me is illegal
(why does bonferroni-correction not work with wilcoxon?)
By ending up with Friedman-test I eliminated problem #2.
But how to deal with problem #1?
Any suggestions?
Thanks,
Sebastian Scala
Veterinary Univeristy Vienna, Austria
before after1
after2
judge# 1 2 3 1 2 3
1 2 3
Horse ID left / right
1 l 1 1 1 1 1
1 1 1 1
r 1 1 1 0 1
0 1 1 1
2 l 0 1 0 0 1
0 1 3 1
r 1 3 0 0 3
0 1 2 0
3 l 2 2 3 2 2
2 2 2 2
r 1 1 2 1 1
2 2 1 2
4 l 3 3 2 3 3
3 0 0 0
r 3 3 3 3 3
3 0 0 0
5 l 2 3 2 2 3
0 2 3 2
r
6 l 1 1 0 0 1
0 1 1 1
r 1 1 0 1 1
0 1 1 0
7 l 3 3 3 3 3
2 3 3 2
r 2 3 1 2 3
1 2 1 3
8 l 0 1 0 1 2
1 1 2 1
r 0 0 0 0 0
0 1 1 0
9 l 1 1 1 1 1
1 1 1 1
r 1 1 1 1 2
2 1 1 1
10 l
r 1 3 1 1 3
0 1 3 1
|