Page 102 - Statistics II for Dummies

86 Part II: Using Different Types of Regression to Make Predictions

Suppose you’re in the marketing department for a major national company
that sells plasma TVs. You want to sell as many TVs as you can, so you want
to figure out which factors play a role in plasma TV sales. In talking with your
advertising people and remembering what you learned in those business
classes in college, you know that one powerful way to get sales is through
advertising. You think of the types of advertising that may be related to sales
of plasma TVs and your team comes up with two ideas:

✓ TV ads: Of course, how better to sell a TV than through a TV ad?
✓ Newspaper sales: Hit ’em on Sunday when they’re reading the paper
before watching the game through squinty eyes that are missing all the
good plays and the terrible calls the referees are making.

By coming up with a list of possible x variables to predict y, you have just
completed step one of a multiple regression analysis, according to the list
in the previous section. Note that all three variables I use in the TV example
are quantitative (the TV ad and newspaper sales variables and the TV sales
response variable), which means you can go ahead and think about a multiple
regression model by using the two types of ads to predict TV sales.

Collecting the Data

Step two in the multiple regression analysis process is to collect the data for
your x and y variables. To do this, make sure that for each individual in the
data set, you collect all the data for that individual at the same time (including
the y-value and all x-values) and keep the data all together for each individual,
preserving any relationships that may exist between the variables. You must
then enter the data into a table format by using Minitab or any other software
package (each column represents a variable and each row represents all the
data from a single individual) to get a glimpse of the data and to organize it
for later analyses.

To continue with the TV sales example from the preceding section, suppose
that you start thinking about all the reams of data you have available to you
regarding the plasma TV industry. You remember working with the advertising
department before to do a media blitz by using, among other things, TV and
newspaper ads. So you have data on these variables from a variety of store
locations. You take a sample of 22 store locations in different parts of the
country and put together the data on how much money was spent on each
type of advertising, along with the plasma TV sales for that location. You can
see the data in Table 5-1.

10_466469-ch05.indd 86 7/24/09 9:32:32 AM