One Between-Subjects Factor: Pairwise Comparisons

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

1

One Between-Subjects Factor: Pairwise Comparisons


An educational psychologist investigated two study strategies: highlighting and
summarization. The materials in the experiment consisted of three versions of a 15-page
passage on operant conditioning. In the first key words and concepts were
highlighted in the text. In the second summarizations were provided at the end of three
major sections in the passage. The third version was the original unmodified passage.
Subjects were randomly assigned to treatments. The criterion variable was scores on a
30-item test. The psychologist is interested in comparing mean performance for
members of each pair of treatments. In the data the unmodified passage was coded TMT
=1, the highlighted passage TMT =2 and the remaining passage TMT =3.
The design is a between-subjects design with one factor. Input the data for this
type of design:

2
Conduct a one-way between-subjects ANOVA. At the top of the out put your will
find the following SPSS syntax:
GET
FI LE=' G: \ 6481\ spss\ mcomp. 1bet \ pc1bet . sav' .
DATASET NAME Dat aSet 2 WI NDOW=FRONT.
UNI ANOVA
TEST BY TMT
/ METHOD = SSTYPE( 3)
/ I NTERCEPT = I NCLUDE
/ PRI NT = DESCRI PTI VE
/ CRI TERI A = ALPHA( . 05)
/ DESI GN = TMT .

Double click in the text. Then highlight the text that begins with UNIANOVA and ends
with a period and copy the text. Then with the SPSS Data Editor displayed, click
File NewSyntax to open the following window:

3
Paste the copied text into the window:

Put the curser in front of the period and make press Enter twice:
4

Type the following code after /DESIGN =TMT
/ LMATRI X = U vs. H TMT 1 - 1 0
/ LMATRI X = U vs. S TMT 1 0 - 1
/ LMATRI X = H vs. S TMT 0 1 - 1

The text in the parentheses is a label for the printout. Typing TMT signals that you are
defining a contrast for the treatment means. The code 1 - 1 0 in the first line says to
compare the first and second treatments. The code 1 0 - 1 in the second line says to
compare the first and third treatments. The code 0 1 - 1 in the third line says to
compare the second and third treatments. Here is what the widow should be:
5

Once the code is typed correctly go to the output and delete it because the new
program will duplicate the output of the original program. If you wish you can save the
syntax so you can reuse it. To run the program, click Run All. The results to test the
contrasts are labeled Custom Hypothesis Tests. Here is an example:

6
Custom Hypothesis Tests #1

Contrast Results (K Matrix)(a)

Dependent
Variable
Contrast
TEST
Contrast Estimate
-3.133
Hypothesized Value
0
Difference (Estimate - Hypothesized)
-3.133
Std. Error
1.212
Sig.
.013
Lower Bound
-5.580
L1
95% Confidence Interval
for Difference
Upper Bound
-.687
a Based on the user-specified contrast coefficients (L') matrix: U vs. H

Test Results
Dependent Variable: TEST
73.633 1 73.633 6.680 .013
462.933 42 11.022
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


L1 refers to the first LMATRIX command:
/ LMATRI X = U vs. H TMT 1 - 1 0
This command generates a test comparing unmodified (TMT =1) highlighted (TMT =2).
To calculate the t statistic to test the null hypothesis of no linear trend, divide the estimate
by the standard error:

3.133
2.58.
1.212
t

= =
Alternatively, you can calculate t as the square root of F: 6.680 2.58. t = =
Here are the results produced by the syntax:
UNI ANOVA
TEST BY TMT
/ METHOD = SSTYPE( 3)
/ I NTERCEPT = I NCLUDE
7
/ PRI NT = DESCRI PTI VE
/ CRI TERI A = ALPHA( . 05)
/ DESI GN = TMT
/ LMATRI X = " U vs. H" TMT 1 - 1 0
/ LMATRI X = " U vs. S" TMT 1 0 - 1
/ LMATRI X = " H vs. S" TMT 0 1 - 1


Univariate Analysis of Variance

[ Dat aSet 3]

Between-Subjects Factors
15
15
15
1
2
3
TMT
N


Descriptive Statistics
Dependent Variable: TEST
19.40 3.542 15
22.53 3.314 15
24.40 3.089 15
22.11 3.857 45
TMT
1
2
3
Total
Mean Std. Deviation N


Tests of Between-Subjects Effects
Dependent Variable: TEST
191.511
a
2 95.756 8.687 .001
22000.556 1 22000.556 1996.018 .000
191.511 2 95.756 8.687 .001
462.933 42 11.022
22655.000 45
654.444 44
Source
Corrected Model
Intercept
TMT
Error
Total
Corrected Total
Type III Sum
of Squares df Mean Square F Sig.
R Squared =.293 (Adjusted R Squared =.259)
a.


8
Custom Hypothesis Tests Index
LMATRIX
Subcom
mand 1:
U vs. H
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 2:
U vs. S
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 3:
H vs. S
Identity
Matrix
Zero
Matrix
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
1
2
3


Custom Hypothesis Tests #1

Contrast Results (K Matrix)
a
-3.133
0
-3.133
1.212
.013
-5.580
-.687
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
TEST
Depende
nt
Variable
Based on the user-specified contrast coefficients (L') matrix: U vs.
H
a.


9
Test Results
Dependent Variable: TEST
73.633 1 73.633 6.680 .013
462.933 42 11.022
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #2

Contrast Results (K Matrix)
a
-5.000
0
-5.000
1.212
.000
-7.446
-2.554
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
TEST
Depende
nt
Variable
Based on the user-specified contrast coefficients (L') matrix: U vs. S
a.


Test Results
Dependent Variable: TEST
187.500 1 187.500 17.011 .000
462.933 42 11.022
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #3

10
Contrast Results (K Matrix)
a
-1.867
0
-1.867
1.212
.131
-4.313
.580
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
TEST
Depende
nt
Variable
Based on the user-specified contrast coefficients (L') matrix: H vs. S
a.


Test Results
Dependent Variable: TEST
26.133 1 26.133 2.371 .131
462.933 42 11.022
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


.



11
One Between-Subjects Factor: Orthogonal Polynomial Analysis
Subjects are randomly assigned to practice a psychomotor task for one to six
trials. Following completion of the practice trials, the subjects complete a criterion trial
on the task. For each number of practice trials, criterion trial data are available for four
subjects. The independent variable is the number of trials. The dependent variable is an
accuracy score on the criterion trial. Higher scores indicate greater accuracy.
The design is a between-subjects design with one factor. Input the data for this
type of design:

12
Follow the directions for the pairwise comparisons. Here is what the first syntax
widow will look like


Type the following code after /DESIGN =TRIALS
/LMATRIX = l i near TRI ALS - 5 - 3 - 1 1 3 5
/LMATRIX = quadr at i c TRI ALS 5 - 1 - 4 - 4 - 1 5
/LMATRIX = cubi c TRI ALS - 5 7 4 - 4 - 7 5
/LMATRIX = quar t i c TRI ALS 1 - 3 2 2 - 3 1
/LMATRIX = qui nt i c TRI ALS - 1 5 - 10 10 - 5 1
13

Once the code is typed correctly go to the output and delete because the new program
will duplicate the output of the original program. If you wish you can save the syntax so
you can reuse it. To run the program, click Run All. The results to test the contrasts
are labeled Custom Hypothesis Tests. Here is an example


Custom Hypothesis Tests #1

14
Contrast Results (K Matrix)
a
-286.750
0
-286.750
42.183
.000
-375.373
-198.127
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix: linear
a.

Test Results
Dependent Variable: ACCURACY
4698.604 1 4698.604 46.209 .000
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.

L1 refers to the first LMATRIX command:
/LMATRIX = l i near TRI ALS - 5 - 3 - 1 1 3 5
This command generates a test of the linear trend. To calculate the t statistic to test the
null hypothesis of no linear trend, divide the estimate by the standard error:

286.750
6.80.
42.183
t

= =
Alternatively you can calculate 46.208 6.80. t = =
Here are the results produced by the syntax:

Univariate Analysis of Variance

15
Between-Subjects Factors
4
4
4
4
4
4
1
2
3
4
5
6
TRIALS
N


Descriptive Statistics
Dependent Variable: ACCURACY
97.75 8.770 4
64.50 4.655 4
60.25 16.820 4
53.50 12.179 4
52.00 6.976 4
49.25 5.620 4
62.88 18.997 24
TRIALS
1
2
3
4
5
6
Total
Mean Std. Deviation N


Tests of Between-Subjects Effects
Dependent Variable: ACCURACY
6470.375
a
5 1294.075 12.727 .000
94878.375 1 94878.375 933.102 .000
6470.375 5 1294.075 12.727 .000
1830.250 18 101.681
103179.000 24
8300.625 23
Source
Corrected Model
Intercept
TRIALS
Error
Total
Corrected Total
Type III Sum
of Squares df Mean Square F Sig.
R Squared =.780 (Adjusted R Squared =.718)
a.


16
Custom Hypothesis Tests Index
LMATRIX
Subcom
mand 1:
linear
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 2:
quadratic
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 3:
cubic
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 4:
quartic
Identity
Matrix
Zero
Matrix
LMATRIX
Subcom
mand 5:
quintic
Identity
Matrix
Zero
Matrix
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
Contrast Coefficients
(L' Matrix)
Transformation
Coefficients (M Matrix)
Contrast Results (K
Matrix)
1
2
3
4
5


Custom Hypothesis Tests #1

17
Contrast Results (K Matrix)
a
-286.750
0
-286.750
42.183
.000
-375.373
-198.127
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix: linear
a.


Test Results
Dependent Variable: ACCURACY
4698.604 1 4698.604 46.209 .000
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #2

Contrast Results (K Matrix)
a
163.500
0
163.500
46.209
.002
66.418
260.582
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix:
quadratic
a.


18
Test Results
Dependent Variable: ACCURACY
1272.964 1 1272.964 12.519 .002
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #3

Contrast Results (K Matrix)
a
-128.000
0
-128.000
67.643
.075
-270.113
14.113
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix: cubic
a.


Test Results
Dependent Variable: ACCURACY
364.089 1 364.089 3.581 .075
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #4

19
Contrast Results (K Matrix)
a
25.000
0
25.000
26.679
.361
-31.050
81.050
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix: quartic
a.


Test Results
Dependent Variable: ACCURACY
89.286 1 89.286 .878 .361
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


Custom Hypothesis Tests #5

Contrast Results (K Matrix)
a
-53.500
0
-53.500
80.037
.512
-221.651
114.651
Contrast Estimate
Hypothesized Value
Difference (Estimate - Hypothesized)
Std. Error
Sig.
Lower Bound
Upper Bound
95% Confidence Interval
for Difference
Contrast
L1
ACCURACY
Dependent
Variable
Based on the user-specified contrast coefficients (L') matrix: quintic
a.


20
Test Results
Dependent Variable: ACCURACY
45.433 1 45.433 .447 .512
1830.250 18 101.681
Source
Contrast
Error
Sum of
Squares df Mean Square F Sig.


.

You might also like