SPSS for One-Way ANOVA

Example: Evaluation of training programs.

Goal: To see if there is significant difference in learning time using different training methods.

To perform one-way ANOVA, for the data listed in the data table which contain 4 independent
random samples:
1. Enter the dependent variable values and the independent variable (factor variable) values in the
Date Editor. In the SPSS Data Editor sheet, it contains a data sheet for a one-way layout design
with four treatment groups. The data in the following picture were scores from four treatment
groups. Method is the factor variable and learning time is the dependent variable.

2. Click through the following menu selection: Data table: 3 observations in

Analyze / Compare Means / One-Way ANOVA. each sample.
Method Method Method Method
1 2 3 4
10 11 13 18

9 16 8 23

5 9 9 25

Method is a treatment,
group or independent
variable. Learning time is
the dependent variable.

3. Select the dependent or response variable and put into the Dependent List box, and put the
method or treatment variable into Factor box.

1
 4. Click Options button, check Descriptive and Homogeneity-of-Variance box, and click
Continue and click OK.

SPSS Output:
Descriptives

Learning Time
95% Confidence Interval for
Mean
N Mean Std. Deviation Std. Error Lower Bound Upper Bound Minimum Maximum
M1 3 8.00 2.65 1.53 1.43 14.57 5 10
M2 3 12.00 3.61 2.08 3.04 20.96 9 16
M3 3 10.00 2.65 1.53 3.43 16.57 8 13
M4 3 22.00 3.61 2.08 13.04 30.96 18 25
Total 12 13.00 6.24 1.80 9.04 16.96 5 25

Test of Homogeneity of Variances

Learning Time
Levene
Statistic df1 df2 Sig. p-value indicating equal variances
.267 3 8 .848

ANOVA

Learning Time
Sum of
p-value indicating
Squares df Mean Square F Sig. significant
Between Groups 348.000 3 116.000 11.600 .003
difference between
Within Groups 80.000 8 10.000
Total 428.000 11 treatment groups.

Error bar chart using SPSS graph Side-by-side boxplot using SPSS Explore

40 30

30

20
95% CI SCORE

Learning Time

20

10
10

-10 0
N= 3 3 3 3 N= 3 3 3 3
1.00 2.00 3.00 4.00 M1 M2 M3 M4

METHOD Training Method

Both error bar chart and side-by-side box plot above seem to suggest that the treatment group
“four” may be significantly different from treatments 1, 2 and 3.

2
5. To perform multiple comparisons, in the ANOVA dialog box, click the Post Hoc… button and
check Tukey or any other method and click Continue and OK.

SPSS produces two tables. The multiple comparisons table containing confidence intervals can
help us to understand the difference between each pairs of means. If interval doesn’t cover zero, it
implies that the difference between the pair of means are statistically significant.
Multiple Comparisons

Dependent Variable: Learning Time

Tukey HSD

Mean
Difference 95% Confidence Interval
(I) Training Method (J) Training Method (I-J) Std. Error Sig. Lower Bound Upper Bound
M1 M2 -4.00 2.58 .455 -12.27 4.27
M3 -2.00 2.58 .864 -10.27 6.27
M4 -14.00* 2.58 .003 -22.27 -5.73
M2 M1 4.00 2.58 .455 -4.27 12.27
M3 2.00 2.58 .864 -6.27 10.27
M4 -10.00* 2.58 .020 -18.27 -1.73
M3 M1 2.00 2.58 .864 -6.27 10.27
M2 -2.00 2.58 .864 -10.27 6.27
M4 -12.00* 2.58 .007 -20.27 -3.73
M4 M1 14.00* 2.58 .003 5.73 22.27
M2 10.00* 2.58 .020 1.73 18.27
M3 12.00* 2.58 .007 3.73 20.27
*. The mean difference is significant at the .05 level.

The homogenous subsets table can help us to divide the four groups into homogenous
subgroups. Within each subgroup the difference in means is statistically insignificant.
The difference between average learning time of Methods 1, 2 and 3 are statistically
insignificant and their means are significantly different from the mean from Method 4.

Learning Time
a
Tukey HSD
Subset for alpha = .05
Training Method N 1 2
M1 3 8.00
M3 3 10.00
M2 3 12.00
M4 3 22.00
Sig. .455 1.000
Means for groups in homogeneous subsets are displayed.
a. Uses Harmonic Mean Sample Size = 3.000.