Database Reference
In-Depth Information
task and the shipping task? Or between the “My Bag” task and the credit card task?
It's Monday, right before lunch. You've got to start building your PowerPoint presen-
tation, and you need to know what these ease-of-use ratings mean. Within-subjects
ANOVA to the rescue!
7.3
COMPARING SEVERAL MEANS WITH A WITHIN-
SUBJECTS DESIGN
In this scenario, you decide to test whether the mean rating of ease-of-use is the same
or different for the four different tasks. (The subsequent S-N-K test will specii-
cally compare the credit card task and the shipping task, and will also compare the
“My Bag” task and the credit card task.) We would formulate the null and alternate
hypotheses accordingly:
•
H0: The true mean rating of ease-of-use is the same for the four tasks.
•
H1: The true average rating of ease-of-use is
not
the same for the four tasks.
Following the earlier chapters' notation, we might write the hypotheses
2
in a
more formal way:
H0: μ1
=
μ2
=
μ3
=
μ4
H1: The four values of μ are not the same for the four designs.
7.3.1
THE KEY
We'll let you in on a little secret:
when we are studying one factor (e.g., different
designs, or different tasks, etc.) using the type of within-subject design (each subject
is evaluating all tasks), we “make believe” that we are studying TWO factors
. The
irst would be the primary factor under study—the design, or the task, or whatever,
and the second factor is the person or respondent!! That is, we view the situation
as having a template such as that in
Table 7.3
(with, let's say, four ratings with six
people with each “X” being a data value on the 1-5 Likert scale.)
So, while there is one factor that is of primary interest, the “column factor,”
task,
there is also a “row factor”—
respondent
—with each of the six rows representing a dif-
ferent person. Typically, the people comprising the sample are a somewhat random set
of people, but who are nevertheless from the same target audience; in this case, women
from 18-55 years of age with well-above-average disposable income.
SIDEBAR: “REAL FACTORS” OR NOT?
In this chapter, we do not dive into a discussion of some interesting issues that arise when there
are two or more “real” factors being studied. We save that for the next chapter, Chapter 8, in which
we will study the effect of two “real” factors, such as
age
and
gender
(e.g., are there differences in
ease-of-use rating depending on age
and
gender?)—in a study that, indeed, combines two factors in
one study—not unlike what is depicted in
Table 7.3
, except that the second factor (the “row factor”)
would be the age or gender factor). It will be called a “two-factor ANOVA.”
2
Note that the hypotheses above are the same whether we have independent samples (as in Chapter 6)
or repeated measures—a subject evaluates all four designs.
Search WWH ::
Custom Search