The Marketing Book 5th Edition

214 The Marketing Book

Table 9.7 Regression, automatic interaction detection and discriminant

analysis – a comparison

Method Based on Marketing

applications

Main advantages Main limitations

Regression

analysis

Developing a

function

expressing the

association (or

relationship)

between

dependent and

independent

variables

For segmentation,

consumer

behaviour analysis,

sales forecasting

(Speed, 1994)

Enables

predictions about

a dependent

variable (say, sales

figures). Provides

measures of

association

between

independent

variables and

certain important

marketing

dependent

variables

Requires fitting a

regression line and

determining the

parameters. This

could be quite

complex and lead

to certain errors

Automatic

interaction

detection

A computer-based

sequential routine

attempting to

classify objects

into groups as

possible, by

minimizing the

within-group sum

of squares

For market

segments analysis,

assess the effects

of advertising on

retail sales, predict

brand loyalty sales

prediction, etc.

Suitable for

identifying the

different variables

affecting market

segments;

determining the

importance of

each independent

variable and the

form in which it

affects the

dependent variable

Less powerful than

regression.

Minimum group

size should be no

less than 30, and

the original sample

size should be

quite large

Discriminant

analysis

Maximize the ratio

of variance

between group

means, not within-

group variance

Predicting brand

loyalty, consumer

innovators, like/

dislike of a service

(or product), etc.

Enables

predictions of

dependent

variables

Identifying the

statistical

significance of the

discriminant

function; multiple

discriminant

analysis requires a

computer program