Database Reference
In-Depth Information
FIGURE 11.11
Augmenting the data by the predicted probability of a “1” for each row; SPSS with illustra-
tive example.
example, all the values from 3 through 8 are in the data set. And, for X = 3, the
probability of a “1” is very small (0.0072) and for X = 8 quite high (0.95023), and it
would not seem that fruitful to worry about what happens when X = 2 or 9. However,
for the sake of introducing the technique, suppose that we decide that we do wish to
know the predicted probability when X = 9. What do we do?
Obviously, there is no data value with X = 9. Here's the trick. We enter the 9 for
X as an additional data value,
but do not enter any value of Y
. We see this in
Figure
11.12
; see the arrow where we added the “9.” The one thing we know is that the
resulting predicted probability will be higher than the 0.95203 for X = 8.
When we now click on the “Save” option, and then click on “Probabilities” and
then “OK,” we arrive at
Figure 11.13
, which not only gives us the predicted prob-
ability of a “1” for the actual X data values, but also gives us the answer for X = 9
(see arrow in
Figure 11.13
).
In other words, since we added an X value, but did not enter a corresponding Y
value, the logistic regression analysis ignores the “9” for purposes of doing the analy-
sis (and hence, the resulting equation, etc., do not change), but the analysis does give
us a predicted probability of getting a “1,” since this predicted value does not depend
on the speciic Y value.
You can see in
Figure 11.13
that our predicted probability for X = 9 is 0.99163.
By the way, we wish to mention again that if you take a quick look way back at
Figure 11.1
, you can see that at the “end points” of the graph (left or right!!), the rise
Search WWH ::
Custom Search