Will anybody buy? Logistic regression - Improving the User Experience through Practical Data Analytics - page 292

Table 11.2 Questions and Coding in the CharlestonGlobe.com Survey—cont'd

Question Measured Coded Comments

11. Strong sports
content

Same Likert scale 1-5 as checked off

12. Strong arts
content

Same Likert scale 1-5 as checked off

13. What is the
maximum you
would be willing to
pay on a monthly
basis for access to
CharlestonGlobe.
com?

Eight categories, in
dollars:
0, 1-4, 5-9, 10-15,
16-20, 21-25,
26-30, >30

None of the catego-
ries above 1-4 had a
suficient frequency
to stand alone, nor
did they in total. So,
0 was coded 0, and
anything >0 was
coded 1.

14. Is there any kind
of online news experi-
ence that you would
be willing to pay for?

Yes/No Yes = 1

No = 0

1 Further discussion of scale types and interval-scale assumptions are beyond the scope of the
text. Further elaboration can be found in any Marketing Research text.

However, before running and analyzing the regression, we irst consider a couple
of issues of interest to Denise Dangle and others at The Charleston Globe .
First, let's examine the results of question 13: “What is the maximum you would
be willing to pay on a monthly basis for access to CharlestonGlobe.com ?” The fre-
quency distribution for that question is shown in Table 11.3 .
We can see in Table 11.3 that the frequency of responses above $4/month is quite
sparse, and, indeed, none of the categories above $4 has suficient frequency to be a
“stand-alone category.” In fact, if we add them all up (i.e., from third through eighth cat-
egories), we ind a total of only 21 responses. A rule of thumb (from database marketing
studies) for a stand-alone category is to have at least 25 people. The reason for this is
that for a categorical variable, if the frequency is too small, the results you get may be
unreliable. Therefore, we coded the variable for regression-analysis purposes at 0 and
1, where “0” indicates the person is willing to pay nothing for CharlestonGlobe.com,
and a “1” indicates the person is willing to pay something for CharlestonGlobe.com.
Next, we consider the relationship of great interest to Nick Bonovich: the pos-
sible correlation between the importance of Strong Credibility and Willingness
to pay something for an online news experience . We performed a chi-square test
of independence with the two variables. The results are shown in Figure 11.15 .
Nick was right! The chi-square results indicate a strong relationship between the
variables—indeed, the p -value is 0.000 (see horizontal arrow in Figure 11.15 ). A care-
ful look at the frequencies (top section of output) indicates that the direction of the
relationship is that “the more impactful the issue of strong credibility, the more likely
to be willing to pay for an online news experience.” In addition, we can also see from