Can you relate in multiple ways? Multiple linear regression and stepwise regression - Improving the User Experience through Practical Data Analytics - page 252

Table 10.3 The 15 Search Engine Components (as depicted in Chapter 9)
1. Ability to search by job title
2. Ability to search by years of experience
3. Ability to search by location
4. Ability to search by schools attended
5. Ability to search candidates by date of updated resume
6. Ability to search candidates by level of education
7. Ability to search by skills
8. Ability to search candidates by average length of employment at each company
9. Ability to search candidates by maximum salary
10. Ability to search candidates by job type he/she is looking for: full time, part time,
temporary/contract, per diem, intern
11. Ability to search candidates by companies in which they have worked
12. Ability to search candidates by willingness to travel. (Expressed as “no travel ability
required,” “up to 25%,” “up to 50%,” “up to 75%,” “up to 100%”)
13. Ability to search candidates by willingness to relocate
14. Ability to search candidates by security clearance. (Active Conidential, Inactive
Conidential, Active Secret, Inactive Secret, Active Top Secret, Inactive Top Secret,
Active Secret/SCI, Inactive Top Secret/SCI)
15. Ability to perform a Boolean search
engine is available to you at no cost to ind qualiied candidates using the candidate
databases you currently employ. What is your likelihood of adopting this candidate
search engine on a scale of 1-5, where 1 = not at all likely and 5 = extremely likely?”)
Now, let us see which of the 15 variables seem to be important in inluencing a
responder's likelihood to adopt the search engine, and how much, overall, the 15
variables tell us about the responder's likelihood of adoption of the search engine.
Figure 10.8 displays the 16 variables in the study (15 X's and the Y), and the irst
10 data points (11 rows, including the label row) out of the 180 data points in total in
an Excel spreadsheet. The names of the variables have, in some cases, been abbrevi-
ated to save space. The full name of each variable is in Table 10.3 .
To run the full multiple regression using Excel, we do what we have done before,
with labels box checked, the Y range being P1:P181, and the X range being A1:O181.
The output is in Figure 10.9 .
Let us now discuss this output. (We purposely left off a few columns on the right-hand
side of the output that had to do with conidence intervals for the coeficients, so we could
show you the more important parts of the output on one page with a bigger font size.)
First of all, you can note that the multiple r 2 = 0.493 (see horizontal arrow in
Figure 10.9 ). So, all of the variables together explain about 49.3% of the differences
in Y from responder to responder.
We might next note that the F -test is highly signiicant. The p -value is 1.37E-17
(see circle around the p -value of the F -test), which, as we noted earlier, indicates
1.37*10 −17 ; this, of course, is very close to zero.