Will anybody buy? Logistic regression - Improving the User Experience through Practical Data Analytics - page 301

Surprisingly, “Local” is not signiicant in the above regression analysis, although
it is fairly close ( p -value = 0.110; see dashed arrow).
Now compare the top two sections of this regression with the top two sections of
the last step of the stepwise regression analysis depicted in Figure 11.18 .
If we go up to the second section of output, we can see that in the last step of
the stepwise regression analysis, we have 85.2% of the people correctly classiied.
In the multiple regression with the ive original signiicant variables, this value is
84.2% (see arrows in Figures 11.19 and 11.18 ). This is a somewhat negligible dif-
ference, especially when we remember that these values are estimates based on 203
people, and not exact values we would get with the entire population of people that
CharlestonGlobe.com can solicit. Both of these percentage values are only margin-
ally below what we get using all 13 values ( Figure 11.16 ), 87.7%.
If we move up to the top section of output, we see the same pattern. In the stepwise
regression, we have the two pseudo- R -square values of 0.492 and 0.661 for step 5, and
0.487 and 0.654 in the multiple regression with the original ive signiicant variables.
Clearly, there is not much difference between the sets of values, and both are margin-
ally below the 0.523 and 0.703 we obtain in Figure 11.16 using all 13 variables.
What's the takeaway? Ultimately, very similar conclusions are to be reached from
the original multiple regression and the stepwise regression. However, the stepwise
regression has decreased somewhat the importance of Unique Local Content Not
Available on a Local Newspaper Site .
11.6 IMPLICATIONS OF THE SURVEY-DATA ANALYSIS
RESULTS—BACK TO CHARLESTONGLOBE.COM
Looking over the inal results from your stepwise regression (step 5 of Figure 11.18 ),
we can revise our main conclusions that were garnered as a result of the original
binary logistic model. Here are the revised (and inal) conclusions:
1. A person with higher educational level is more likely to be willing to pay for an
online news experience.
2. A person who subscribes to a print newspaper is more likely to be willing to pay
for an online news experience than a person who does not subscribe to a print
newspaper.
3. A person who is of the opinion that having Strong Credibility is more impact-
ful in making a decision about his/her likelihood of paying for an online news
experience is more likely to be willing to pay for online news experience.
4. A person who is of the opinion that having In-Depth Analysis of Business Stories is
more impactful in making a decision about his/her likelihood of paying for an online
news experience is more likely to be willing to pay for an online news experience.
5. A person who is of the opinion that having In-Depth Analysis of National Sto-
ries is more impactful in making a decision about his/her likelihood of paying
for an online news experience is less likely to be willing to pay for an online
news experience.