
Market
Segment Classification Modelling
With Logistic Regression Bruce Ratner, Ph.D. Logistic regression
analysis is a recognized technique for classifying individuals into two
groups. Perhaps less known but equally important, polychotomous
logistic regression (PLR) analysis is another method for performing
classification. The purpose of this article is to present PLR
analysis as a multigroup classification technique. I illustrate
the technique using a cellular phone market segmentation study to build
a market segment classification model as part of a customer
relationship management strategy better known as CRM.
I start the discussion
by defining the typical twogroup (or binary) logistic regression
model. After introducing necessary notation for expanding the binary
logistic regression model, I define the PLR model. For readers
uncomfortable with such notation, the PLR model provides several
equations for classifying individuals into one of many groups. The
number of equations is one less than the number of groups. Each
equation looks like the binary logistic regression model.
After a brief review of
the estimation and modeling processes used in polychotomous logistic
regression, I illustrate PLR analysis as a multigroup classification
technique along with CHAID
in a case study based on a survey of
cellular phone users. The survey data was used initially to segment the
cellular phone market into four groups. I use PLR analysis to build a
model for classifying cellular users into one of the four groups.

