Database Reference
In-Depth Information
“Accidenti! Only 3 out 10 could enter their credit card info?” Ranieri gasps.
“Yes and only 2 out of 10 could use the shipping method dropdown correctly.”
The CEO turns to the Creative Director. “It looks like we have some work to do.
What problem do we tackle irst?”
“Well, when I looked at the posttest ease-of-use ratings between the credit
card problem and shipping problem, the shipping problem seemed more serious.
The mean rating of ease-of-use for the credit card was 1.8 compared to 3.1 for
the credit card problem. I ran an ANOVA and Newman-Keuls test to verify the
difference.
“A what? The CEO asks. “I take it that's stats stuff, right?”
“Yes, in a case like this you use a within-subjects ANOVA to determine if
there are statistical differences in the irst place among the multiple means, then
a Newman-Keuls test to determine which means are speciically different from
others—i.e., determine the way in which the means differ. In this case, all four rat-
ings' means are signiicantly different from one another.”
“English, please?” the CEO says with a laugh. He's warming to you.
“As bad as the credit card problem is, the shipping method problem is likely
worse. I'd work on that irst. To double check, we should look at our Google Analyt-
ics data and see if we're seeing big abandonment in the checkout cart.”
The CEO looks at the Creative Director: “Well at least we've got a plan. You're
on it?”
“We're on it!”
On the way back to your desk, the Creative Director seems pleased.
“Nice work. That was tough news to report. But you had the goods.”
“Thanks. Is he always that grumpy?”
“Are you kidding?” the Creative Director asks. “He was in a great mood, consid-
ering. That's the irst time I ever saw him laugh!”
7.9 SUMMARY
In this chapter we introduced within-subject designs (or “repeated measures”
designs.) The classic description of such a design is to have a factor under study,
with several levels (more than two), but have the same person rating the ease-of-use
of each task. This is in contrast to the case of having independent samples, in which
each person evaluates or rates only one entity—in essence, each person provides only
one data value. We illustrated the within-subjects design by an example in which we
studied one factor (task) having four levels (i.e., four different tasks). Each person
provided four data values, one for each task under investigation. The measure (or
output variable or dependent variable) was ease-of-use evaluation.
We saw that the procedure we used for analysis was ANOVA, even though we
needed to be careful to specify what was necessary in telling SPSS what to do, in com-
parison to the case of independent samples, the latter introduced in Chapter 6. After
seeing the results of the analysis, we carried out the S-N-K test introduced in Chapter 6.
 
Search WWH ::




Custom Search