Discriminant Analysis

Objective of Discriminant analysis

The goal of this method is the analysis of group differences. The basis for this is a sample which can be unambiguously broken down into groups. Information relating to a series of attributes must be present for each member of this sample.

One question addressed by this analysis is whether the groups differ significantly in terms of these attributes. The analysis shows which attributes are particularly suitable for distinguishing the groups. The groups can be described and positioned in terms of these attributes.

Furthermore, this method makes it possible to classify persons whose group membership is unknown. Thus, for instance, in mailing campaigns it becomes possible to use existing data in order to target exactly those persons who are highly likely to respond. Also, when the attribute information is collected in the screening questionnaire, survey participants can be assigned to their appropriate quota segments (clusters).

Prerequisites

The descriptive attributes must be scaled metrically. There must be at least twice as many cases with known group membership available as the number of attribute variables.

The number of groups should not exceed the number of attributes.

If cases with unknown group membership are to be classified (e.g. through a screening interview), it is important that these originate from the same population as the cases used in the analysis.

Contact Discriminant Analysis

Prof. Dr. Heiko Schimmelpfennig
+ 49 40 25 17 13 35
hschimmelpfennig@ifad.de