1
SPSS Chapter 12 Example 1 - One-Way Analysis of Variance (ANOVA)
A study of reading comprehension in children compared three methods of
instruction, known as basal, DRTA, and strategies (denoted as “Strat” in this
example). A measure of reading comprehension (the “comp” variable seen below)
was received after the instruction was completed. We are interested in comparing
the reading comprehension of the three instruction groups. We are testing
H0: µB=µD=µS
Ha: µB≠µD≠µS
After opening the file, the data appear in the SPSS Data Editor window just like
the following (please note that for the variable entitled Group, basal=1, DRTA=2
and strategies=3):
Dr. Robert Gebotys 2006
2
Follow these steps to perform a One-Way ANOVA:
1. Click Analyze, click Compare Means, and click One-Way ANOVA. The
following window will appear.
2. Click “comp” (a.k.a. “comprehension score”) and click!to move “comp”
into the box entitled Dependent List.
3. Click “group” and click!to move “group” into the box entitled Factor.
Dr. Robert Gebotys 2006
3
4. To calculate contrasts, click the button entitled Contrasts and the following
window will appear.
5. We are interested in two contrasts: basal vs. DRTA and strategies (-2, 1, 1) as
well as DRTA vs. strategies (0, 1, -1). The coefficients of each contrast are
entered separately in the box entitled Coefficients. After the first coefficient is
entered (i.e., -2), click Add. Enter the remaining coefficients of the first
contrast (i.e., 1 and 1) in the same manner. Click Next to enter the second
contrast.
6. Repeat step 5 for the second contrast, then click Continue.
Dr. Robert Gebotys 2006
4
7. To calculate post hoc multiple comparisons, click the button entitled Post Hoc
and the following window will appear.
8. Click LSD and Bonferroni so that a checkmark (!) appears in the boxes before
those multiple comparisons. Click Continue.
9. Click OK.
Dr. Robert Gebotys 2006
5
The SPSS output for this example of the One-Way ANOVA is the following:
ANOVA
comp
Sum of Mean
Squares df Square F Sig.
Between Groups 357.303 2 178.652 4.481 .015
Within Groups 2511.682 63 39.868
Total 2868.985 65
The null hypothesis of equal means is rejected. F(2,63)=4.481,p=.[Link]
researcher knows that there is at least one difference among the means. Pre-
planned comparisons which are orthogonal can also be tested. See the notes for an
interpretation of each contrast.
Contrast Coefficients
Group
Contrast basal DRTA strat
1 -2 1 1
2 0 1 -1
The orthogonal comparisons are tested in the table below. Under the null
hypothesis the contrast=0, the alternative hypothesis indicates the contrast is not
equal 0. Contrast one is significant p=.009 however contrast two is not significant
p=.202. The average of the d and s groups is different from the b group. The d and
s groups do not differ.
Contrast Tests
Value of Std. Sig.
Contrast Contrast Error t df (2-tailed)
comp Assume equal variances 1 4.455 1.649 2.702 63 .009
2 2.455 1.904 1.289 63 .202
Does not assume equal 1 4.455 1.563 2.851 47.945 .006
variances 2 2.455 1.998 1.228 39.661 .227
The multiple comparisons are given on the next table. The tests both
indicate that the b and d group means differ.
Dr. Robert Gebotys 2006
6
Multiple Comparisons
Dependent Variable: comp
95% Confidence
Mean Interval
(I) GROUP (J) GROUP Difference Std. Lower Upper
(I-J) Error Sig. Bound Bound
LSD basal DRTA -5.682* 1.904 .004 -9.486 -1.877
strat -3.227 1.904 .095 -7.032 .577
DRTA basal 5.682* 1.904 .004 1.877 9.486
strat 2.455 1.904 .202 -1.350 6.259
strat basal 3.227 1.904 .095 -.577 7.032
DRTA -2.455 1.904 .202 -6.259 1.350
Bonferroni basal DRTA 5.682* 1.904 .012 -10.364 -.999
strat -3.227 1.904 .285 -7.910 1.455
DRTA basal 5.682* 1.904 .012 .999 10.364
strat 2.455 1.904 .606 -2.228 7.137
strat basal 3.227 1.904 .285 -1.455 7.910
DRTA -2.455 1.904 .606 -7.137 2.228
* The mean difference is significant at the .05 level.
Dr. Robert Gebotys 2006