Scribd
Upload
210 views1 page

Mult Comp

This document summarizes several important multiple comparisons procedures for analyzing differences between means from multiple treatment groups: 1) Least Significant Difference (LSD) compares means to a threshold based on a t-distribution. 2) Sidak correction provides a tighter bound than Bonferroni by adjusting the critical value of significance. 3) Stepdown Bonferroni assigns adjusted p-values in order from smallest to largest. 4) Scheffe's method and Tukey's procedure compare means to critical values based on an F-distribution and studentized range statistic respectively to control familywise error rate.

Uploaded by

Héctor Yendis
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
210 views1 page

Mult Comp

This document summarizes several important multiple comparisons procedures for analyzing differences between means from multiple treatment groups: 1) Least Significant Difference (LSD) compares means to a threshold based on a t-distribution. 2) Sidak correction provides a tighter bound than Bonferroni by adjusting the critical value of significance. 3) Stepdown Bonferroni assigns adjusted p-values in order from smallest to largest. 4) Scheffe's method and Tukey's procedure compare means to critical values based on an F-distribution and studentized range statistic respectively to control familywise error rate.

Uploaded by

Héctor Yendis
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Statistics 850 Clayton January 29, 2004

Some Important Multiple Comparisons Procedures


In the following, suppose there are k treatments, let r =
_
k
2
_
, and let t
ii
=
y
i
y
i

2/n
.
Least Signicant Dierence
Compare t
ii
to T
dfErr,/2
. This is equivalent to comparing | y
i
y
i

| to T
dfErr,/2
s

_
2/n.
Sidak
Like Bonferroni, but use CER = 1 (1 )
1/r
. This produces a bound that is slightly tighter than
Bonferroni.
A Stepdown Bonferroni p-value
Put the unadjusted p-values in order from smallest to largest: p
1
< p
2
< < p
r
. The adjusted
p-values are p
1
= min{1, rp
1
}; p
2
= min{1, max{ p
1
, (r 1)p
2
}}; p
3
= min{1, max{ p
2
, (r 2)p
3
}}; etc
Schee
Compare |t
ii
| to
_
(k 1)F
k1,dfErr,
. Equivalently, compare | y
i
y
i

| to
_
(k 1)F
k1,dfErr,
s

_
2/n.
Schees result can be extended to contrasts:
P
_
k

1
c
i

i

k

1
c
i

Y
i

_
(k 1)F
k1,dfErr,
s

c
2
i
/n
i
c s.t. c

1 = 0
_
= 1
Tukey
Compare | y
i
y
i

| to Q
k,dfErr,
s

_
1/n where Q is chosen such that
P
_
max
i,i

|(

Y
i

i
) (

Y
i


i
)|
s

n
Q
_
= 1
This method also works for contrasts: P
_

k
1
c
i

i


k
1
c
i

Y
i
Qs

k
1
|c
i
|/2 c s.t. c

1 = 0
_
=
1
Student-Newman-Keuls
Same as Tukey for means separations, except, for a subset of size m, use Q
m,dfErr,
.
Duncans New Multiple Range Test
Like SNK, but for a group of size m, use Q
m,dfErr,m
where
m
= 1 (1 )
m1
.
Fishers Least Signicant Dierence
Also called a Protected LSD: rst perform the F test in ANOVA, and proceeds with LSD means
comparisons only if the F test is signicant.
Hayter
Like Fishers LSD, rst do the F test. If it rejects H
0
, then use Tukeys method to compare means,
but use Q
k1
instead of Q
k
.
References
Hsu, J. C., 1996, Multiple Comparisons. Theory and Methods
Miller, R. G., 1981, Simultaneous Statistical Inference (Second Edition)
1

You might also like