Will anybody buy? Logistic regression - Improving the User Experience through Practical Data Analytics - page 298

FIGURE 11.17
The logistic regression dialog box with “Method” pulled down; SPSS with CharlestonGlobe.com.
Another difference between the binary logistic multiple regression and binary
logistic stepwise regression results is the difference of the coeficient for the “Sub-
scriber to Printed Newspaper” variable. It's gone down from 17.7 to 10.7 (rounded).
However, the p -value has remained 0.000 (which, we recall means 0.000 to three
digits). Obviously, this variable is still a very strong predictor. Our enthusiasm for
this predictor is still extremely high.
The inal noteworthy point is that the stepwise results have thrown us an interest-
ing curve. You'll notice the “national” coeficient is signiicant ( p -value = 0.033)
and is negative! (Remember, column B displays the slope of the linear equation,
when the Y*c is the LN (natural log) of the odds. When B is negative, it means
that increasing that X will decrease the LN of the odds (Y*c), which will, in turn,
decrease the odds, which, in turn, will decrease the probability of willingness to pay.
Review pages 281 and 282 for how the probability is determined.) This suggests also
that there is a negative impact on willingness to pay for an online news experience