Chapter 7
The Two-Way Layout
Some Changes to pages 9 – 12:
7.4 One-sided Comparisons of Treatments with Control
Based on Friedman Rank Sums (Nemenyi, Wilcoxon – Wilcox, Miller)
Used after rejecting the null hypothesis of no difference between treatment effects. Compares
each treatment with a control group (call it treatment 1).
Hypotheses of interest are
•
Ho: τ
u
= τ
1
vs.
Ha: τ
u
> τ
1
for all u = 2, 3, …, k
with decision Rule:
Decide τ
u
> τ
1
if (R
u
– R
1
) ≥ r*
α
.
•
Or Ha: τ
u
> τ
1
for all u = 2, 3, …,k
with decision rule
τ
1
< τ
u
if (R
1
– R
u
) ≥ r*
α
.
Example:
Use the hydrobromide data of the example on page 3 (copied below) to test whether
any of the hydrobromides produce
a higher average number of hours of sleep per night that
does the “no hypnotic,” i.e., treatment A.
Patient
A
B
C
D
1
0.6
(1)
1.3
(2)
2.5
(4)
2.1
(3)
2
1.1
(1.5)
1.1
(1.5)
5.7
(3)
5.8
(4)
3
2.5
(1)
6.2
(2)
8
(3)
8.2
(4)
4
2.8
(1)
3.6
(2)
4.4
(4)
4.3
(3)
5
2.9
(1)
4.9
(2)
6.3
(3)
6.4
(4)
6
3.0
(2)
1.4
(1)
3.8
(3)
4.4
(4)
R
j
(7.5)
(10.5)
(20)
(22)
We are interested in testing
Ho: τ
u
= τ
1
vs. Ha: τ
u
> τ
1
,
for all u = 2, 3, 4
simultaneously.
From Table A25, with n = 6, k = 4 and α = 0.0404, we find
Hence we will decide
τ
u
> τ
1
if R
u
– R
1
≥ 10.
Show that,
from the observed data we have
R
1
= 7.5, R
2
= 10.5, R
3
= 20 and R
4
= 22
and hence
Hence we decide that Treatments C and D are better than control (Treatment A), the others are
not different from the control.
This
preview
has intentionally blurred sections.
Sign up to view the full version.

This is the end of the preview.
Sign up
to
access the rest of the document.
- Summer '11
- YESILCAY
-
Click to edit the document details