Академический Документы
Профессиональный Документы
Культура Документы
Post hoc tests in SPSS are available in more than one procedure, including
ONEWAY and GLM.
Notation
The following notation is used throughout this appendix unless otherwise stated:
k
ni
xi
Mean at level i
si
vi
s pp
ni
(n 1)
i
i =1
vi , j
k*
k k 1 / 2
0 5
Appendix 10
Qi, j
s pp
nh
1 1
1
+
2 ni n j
nh =
ni1
1i k
Qh
s pp / nh
Sr ,m =
s m be an
xi , and
; @
is called the Studentized range. The upper- critical point of this distribution is
denoted by S ,r ,m .
The quantity
Mr,m =
max(| x1 |,K,| x r |)
sm
is called the Studentized maximum modulus. The upper- critical point of this
distribution is denoted as M ,r,m .
Methods
The tests are grouped below according to assumptions about sample sizes and
variances.
Equal Variances
The tests in this section are based on the assumption that variances are equal.
Appendix 10
Waller-Duncan t Test
The Waller-Duncan t test statistic is given by
vi, j =| xi x j | t B ( w, F, q, f )S 2 / n
where t B (w, F, q, f ) is the Bayesian t value that depends on w ( a measure of the
relative seriousness of a Type I error versus a Type II error), the F statistic for the
one-way ANOVA,
F=
MStreat
MSerror
and
S 2 = MSerror
Here f = k (n 1) and q = k 1. MSerror and MStreat are the usual mean squares in
the ANOVA table.
Only homogeneous subsets are given for the Waller-Duncan t test. This method is
for equal sample sizes. For unequal sample sizes, the harmonic mean nh is used
instead of n.
Rank the k means in ascending order and denote the ordered means as
x(1) ,K, x( k ) .
2.
Determine the range value, R ,k , f , for the specific test, as shown in Range
Values below.
3.
If x( k ) x(1) > Qh R ,k , f , there is a significant range and the ranges of the two
sets of k 1 means { x(1) ,K, x( k 1) } and { x(2 ) ,K, x( k ) } are compared with
Qh R ,k 1, f . SPSS continues to examine smaller subsets of means as long as
the previous subset has a significant range.
Appendix 10
For some tests, Qi, j is used instead of Qh , as indicated under specific tests
described in Range Values below.
4.
Each time a range proves nonsignificant, the means involved are included in a
single groupa homogeneous subset.
Range Values
Range values for the various types of tests are provided below.
Student-Newman-Keuls (SNK)
R ,r, f = S ,r, f
The confidence intervals of the mean difference are calculated using Qi, j instead of
Qh .
Tukeys b (TUKEYB)
R ,r, f =
S ,r , f + S , k , f
2
R ,r, f = 2( k 1) F1 ( k 1, f )
Appendix 10
The confidence intervals of the mean difference are calculated using Qi, j instead of
Qh .
Hochbergs GT2 (GT2)
R ,r, f = 2 M ,k * , f
The confidence intervals of the mean difference are calculated using Qi, j instead of
Qh .
Gabriels Pairwise Comparisons Test (GABRIEL)
The test statistic and the critical point are as follows:
| xi x j | s pp (
1
1
+
) M , k * , f
2 ni
2n j
(1)
R ,r, f = 2 F1 (1, f )
Appendix 10
R ,r, f = 2 F1 (1, f )
where = / k *
Sidak t Test (SIDAK)
R ,r, f = 2 F1 ,1, f
where = 1 (1 )
2
k ( k 1)
Dunnnett Tests
For the Dunnett tests, confidence intervals are given only for the difference
between the control group and the other groups.
Dunnetts Two-Tailed t Test (DUNNETT)
When a set of new treatments ( xi ) is compared with a control ( x 0 ), Dunnetts twotailed t test is usually used under the equal variances assumption.
For two-tailed tests,
1
1
+
n0 ni
T = max {| Ti |}
1i k
( xi x 0 )
2
and sdd
where Ti =
1
1
s dd
+
n0 ni
ni
i =0
j =1
i =0
( x ij xi. ) 2
(ni 1)
Appendix 10
1
1
+
n0 ni
T = max {Ti }
1i k
Confidence intervals are given only for the difference between the control group
and the other groups.
Dunnetts One-Tailed t Test (DUNNETTR)
This Dunnetts one-tailed t test indicates whether the mean at any level is larger
than a reference category.
xi x 0 < dLk ,v s dd
1
1
+
n0 ni
T = max {Ti }
1i k
Confidence intervals are given only for the difference between the control group
and the other groups.
r =
%&1 (1 )
K'
r/k
if r < k 1
if r k 1
Appendix 10
Note: For homogeneous subsets, the ni and n j are used for the R-E-G-W F test and
the R-E-G-W Q test. To apply these methods, the procedures are same as in
Constructing Homogeneous Subsets on p. 3, using the tests provided below.
Ryan-Einot-Gabriel-Welsch Based on the Studentized Range Test (QREGW)
The R-E-G-W Q test is based on
max {( x i x j ) min(ni , n j )} / s pp S r ,r , f
i, j R
(r 1)s 2pp
v=
( si2 / ni + s 2j / n j )2
si4 / ni2 vi + s 4j / n 2j v j
| xi x j | Qi*, j R ,r,v
Appendix 10
where
Qi*, j =
2
si2 s j
+
ni n j
R ,r,v = S ,k ,v / 2
Tamhanes T2 (T2)
Tamhanes T3 (T3)
R ,r,v = M ,k * ,v
Dunnetts C (C)
R ,r,v =