Information Technology Reference
In-Depth Information
These are the probabilities of
the differences occurring by chance
alone for the pairwise comparisons
for the main effect of drinks.
Pairwise Comparisons
Measure: MEASURE_1
95% Confidence
Interval for
Difference a
Mean
Difference
(I-J)
Lower
Bound
Upper
Bound
(I) DRINKS (J) DRINKS
Sig. a
Std. Error
1
2
3
1
3
1
2
3.083
5.083
3.083
2.000
5.083
2.000
.539
.473
.539
.483
.473
.483
4.987
6.755
1.179
3.707
3.412
.293
*
*
*
*
*
*
.007
.000
.007
.027
.000
.027
1.179
3.412
4.987
.293
6.755
3.707
2
3
Based on estimated marginal means
*. The mean difference is significant at the .05 level.
a. Adjustment for multiple comparisons: Bonferroni.
Figure 11.23
Pairwise comparison tests for the main effect of drinks .
We are thus making the two “horizontal” comparisons with respect to
Figure 11.2 - we are comparing driving errors for the three numbers of
drinks for each vehicle. It appears from the Pairwise Comparisons table
that under the first level of vehicle (cars) all of the conditions differ from
each other but that under the second level of vehicle (SUV) driving errors
increase from zero drinks to one drink but do not further significantly
increase after three drinks.
11.10.2 MAIN EFFECT OF DRINKS
The Paired Comparisons table for the main effect of drinks is shown
in Figure 11.23. Overall, driving errors increased with increasing alcohol
consumed. However, we recognize from just having discussed the inter-
action that this is an overgeneralization since driving errors tapered off
when the students were behind the wheel of an SUV. It should therefore
notbeemphasizedintheresultssection.
11.11 PERFORMING THE OMNIBUS ANALYSIS IN SAS
11.11.1 STRUCTURING THE DATA FILE
The data file in Excel format is shown in Figure 11.24. Note that it is in
univariate or stacked mode as explained in Section 10.15.2. Briefly, each
row can contain no more than a single score on the dependent variable. In
the present 2
3 within-subjects design, each participant was measured
six times, once under each condition. To conform to univariate structure,
we must use six lines for each case, one for each condition.
To illustrate our data structure, consider the participant identified as
subid 1 occupying lines 2 through 7 in the Excel file. Row two (the first
×
Search WWH ::




Custom Search