Will anybody buy? Logistic regression - Improving the User Experience through Practical Data Analytics - page 290

in the curve is not very steep, and that is consistent with the difference from X = 8
to X = 9 being relatively small (roughly 0.96 to 0.99), while the curve rises far more
steeply near the “middle” of the data values, and correspondingly, from X = 5 to
X = 6 engenders a much larger increase (roughly 0.16 to 0.48).

11.5 CHARLESTONGLOBE.COM SURVEY DATA
AND ITS ANALYSIS
You'll recall that in an attempt to determine what kinds of online news content
consumers would be willing to pay for, you launched an online survey. The survey
probed on current news consumption, willingness to pay for online news, and what
attributes of an online news experience would have the most impact of the respon-
dent's willingness to pay for digital news content.
Your survey began with demographic questions of age, gender, income, and level of
education. On a hunch that there might be a correlation between print subscribers and
online use, you added a question about whether the participant has a print subscription.
Then, the meat of the survey: participants were offered seven attributes of online
news content, and each participant was asked to rate each attribute's impact on his/
her willingness to pay for online news content. (Each participant was asked to rate
each one on a scale of 1-5, where 1 = not at all impactful to 5 = extremely impactful.)
The attributes were as follows:
• Strong credibility

• Unique local content not available on local newspaper sites

• In-depth analysis of national stories

• In-depth analysis of international stories

• In-depth analysis of business stories

• Strong sports content

• Strong arts content

You then asked how much a participant was willing to pay per month for online
news, using a dropdown of different denominations. Finally, the survey posed the
crucial question, “Is there any kind of online news experience that you would be
willing to pay for?” as a “yes” or “no” response.
In total, the survey launched with 14 questions. You ran it on various news forums
and communities.
After 1 week, you assess the responses. After some serious cleaning of the data,
you end up with a sample size of 203 usable responses. The questions and response
options and how they were coded are listed in Table 11.2 .
After you imported and recoded your data into SPSS, you're ready to perform the
logistic regression. The irst 20 lines of the 203 rows of data are displayed in Figure 11.14 .
Our primary objective at the moment is to ind out how to predict who is willing
to pay for some kind of news online. In other words, variable 14 (see Table 11.2 or
column 14 (right-hand most column) in Figure 11.14 ) is our Y (dependent variable).
Since it is a yes/no variable, our regression analysis will be binary logistic regression.