We continue the core
textbook example from Chapter 1. Remember that the company is interested
in explaining the first-year sales of its services
offerings. You might generally assume that sales
are at least partially dependent on
(explained by) various aspects including:
-
“License” which
is a categorical variable measured as “Freeware” or
“Premium.”
-
“Size” is
an ordinal variable with levels “Small”, “Medium”
and “Big.”
-
“Trust” refers
to the trust the customer has in your product and company. You originally
measured trust through four questions in an online survey that you
sent to key account holders at the client (measured on a 0 to 100
Likert scale). As per the discussion in Chapter 9, you established
that internal reliability exists for the four items. (You might also
wish to analyze these in terms of factor analysis, which is not covered
in this book). You have therefore averaged the trust items to give
you a single summated score for trust for each customer.
-
“Customer
satisfaction” is also measured through
four questions (this time on a 1 to 7 scale) in a paper-based customer
survey. Like trust, in Chapter 9 we analyzed these four items for
internal reliability, however there we found that the fourth item
had poor reliability. Therefore, to create the final score we averaged
only Satisfaction01-Satisfaction03.
-
“Enquiries” refers
to the average number of enquiries about the core product logged with
the call center or online help by customers, per month, since starting
use of the product. This data is provided by your in-house CRM data
systems.
“Data05_Regression_Initial”
in the book materials shows the dataset for the case example, in this
case isolating only the final aggregated variables used in the regression,
which in total has 279 observations. As can be seen, in its raw format,
the table has both character (word) and numerical data points. Words
have initially been used for the categorical and ordinal variables.
Now that the scene is
set, the following sections discuss the broad ideas behind linear
regression.