Database Reference
In-Depth Information
the variable with the strongest coeficient and highest signiicance. The results are
only meaningful within the context of the entire study about an online news experi-
ence. And other variables came close to signiicant, like unique local content and arts
content. Last, an alpha of 0.05 is pretty rigorous. I could redo the regression with an
alpha of 0.10 and….”
“OK, OK,” Dangle interrupts, motioning you to stop. “You've done your home-
work. And God knows this stats stuff makes my head spin.” She looks directly at
you: “Do me a favor. Trim down your preso to no more than ive pages. I'll schedule
a meeting with Mark.” As she gets up to leave, she offers the irst compliment of your
tenure: “Thanks. Good stuff.”
As Dangle shoots down the hall, Nick comes up and offers his hand in
congratulations.
“Wow. You, my friend, are presenting to CEO Mark Munzer. Spectacular! You
might have just saved CharlestonGlobe.com from the digital graveyard.”
11.7
SUMMARY
In this chapter we introduced binary logistic regression. The key to its use is that the
dependent variable, Y, is a Yes/No-type (i.e., two possibilities) categorical variable. We
discussed how the technique works, and then went through a small example in detail,
using SPSS (Basic Excel does not accommodate binary logistic regression). Next we
addressed the CharlestonGlobe.com data, which had 13 independent variables (called
“covariates” by SPSS), and performed a multiple binary logistic regression analysis,
followed up by a stepwise binary logistic regression analysis (the Method is labeled
“Forward: LR” by SPSS). We then interpreted and discussed the results.
11.8
EXERCISE
1.
Use the SPSS data set from the previous chapter, “Chapter 10..Exercise 1.data,”
to run a binary logistic regression. The dependent variable (unused in the exer-
cise in Chapter 10) is to be the variable labeled “Purchase” (right-hand-most
column in the data set). It represents whether the responder feels that he/she is
more likely than not to purchase the search engine (“1”), or is more likely not to
purchase it (“0”). The output is in a ile, “Chapter 11..Exercise 1.output.”
a.
Which attributes are signiicant?
b.
What percent of the cases, overall, are predicted correctly?
c.
What percent of the “1's” are predicted correctly?
d.
Repeat parts a, b, and c, performing a stepwise regression analysis (using the
“Forward: LR” option).
The solutions are discussed in a Word ile, “Chapter 11.Exercise 1.answers.”
