Information Technology Reference
In-Depth Information
8
Driving a Car
6
Driving an SUV
4
2
0
No Drinks
One Drink
Three Drinks
Number of Alcoholic Drinks
Figure 11.2
A graph of the two-way interaction.
11.4.2 TWO-WAY INTERACTION
The Vehicle
Drinks interaction is statistically significant. Thus, the par-
ticular combinations of the levels of the independent variable are more
predictive of driving errors than either independent variable in isolation.
The interaction is graphed in Figure 11.2. Based on visual inspection of
the figure, it appears that students with no alcohol consumed have very
few driving errors for both vehicles and that driving errors comparably
increase for both vehicles when the students have had one drink. After three
drinks, driving errors continue to increase at about the same rate when
students are behind the wheel of a sports car but appear to increase only
modestly (or perhaps level off) when they are driving an SUV. We need
to perform tests of simple effects to fully explicate this interaction effect
by determining which means are statistically different from which others.
×
11.5 THE ERROR TERMS IN A TWO-WAY
WITHIN-SUBJECTS DESIGN
Each effect of interest in a within-subjects design is associated with its own
error term. This is always true in a within-subjects design. The reason for
this is because each effect interacts with the subjects effect as described
in Chapter 10, and this interaction is “noise” or unpredictability in the
effect across the different participants in the study. This unpredictability
is taken as an error term. Because each effect can interact differently with
subjects, each interaction is keyed to its respective effect.
11.5.1 THE MAIN EFFECT OF VEHICLE
The error term for the main effect of Vehicle is represented by the Vehi-
cle
×
Subjects interaction, shown as Error (
A
×
S
) in the summary table.
