ENBIS: European Network for Business and Industrial Statistics

LATENT CLASS ANALYSIS FOR MARKETING SCALE DEVELOPMENT

21 September 2009, 15:05 – 15:25

Abstract

Submitted by

Francesca Bassi

Authors

Francesca Bassi

Affiliation

Statistics Department, University of Padova, Italy

Abstract

Measurement scales are a crucial instrument for research in marketing in order to measure unobservable variables as attitudes, opinions, beliefs. Examples of unobservable variables related to marketing are customer satisfaction, purchase involvement, brand loyalty, scepticism towards advertising and many other.
In using, evaluating, or developing multi-item scales, a number of guidelines and procedures are recommended to help ensure that the measure is psychometrically as sound as possible. These procedures are outlined in the psychometric literature since the late seventies. Traditionally, with some exceptions, the literature has followed the procedure outlined by Churchill (1979) who identifies a number of steps to take in developing a measure. These steps refer to construct definition and domain and scale validity, reliability, dimensionality, and generalizability. Various statistical instruments are used in the scale developing steps, these almost always refer to variables measured on a metric scale (examples are correlation coefficients, factorial analysis, regression models). Items forming scales are instead almost always measured on a level which is different from the interval one; often items are ordinal, in some rare cases, nominal. Likert ,semantic differential, Staple and Thurstone scales, for example, generate ordinal variables.
In this paper, I show how the implementation of latent class analysis may improve the process of measurement scale development since it explicitly considers that items generate ordinal or even categorical variables. Specifically, applying appropriate latent class models allow to assess scale validity and reliability more soundly than the methods traditionally used, in the first place factorial analysis.

Return to programme