|
Activity Number:
|
524
- Contributed Poster Presentations: Section on Statistics in Marketing
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Marketing
|
|
Abstract #324313
|
|
|
Title:
|
Clustering Methods for Ordered Categorical Data with Response Style
|
|
Author(s):
|
Mariko Takagishi* and Michel van de Velden and Hiroshi Yadohisa
|
|
Companies:
|
Graduate School of Culture and Information Science, Doshisha University and Department of Econometrics, Erasmus School of Economics, Erasmus University Rotterdam and Department of Culture and Information Science, Doshisha University
|
|
Keywords:
|
ordered categorical data ;
clustering ;
spline ;
response style ;
Likert scales
|
|
Abstract:
|
Questionnaire-based surveys often contain ordered categorical data such as those based on Likert scales. Furthermore, clustering individuals based on survey items is useful for discovering latent structures. However, cluster analysis of ordered scale data may be affected by response styles. That is, an individual's systematic response tendency that is independent of item contents. For example, "extreme response style" is a tendency to choose categories at the ends of the scale. A cluster of individuals with "extreme response style" can be mistakenly identified as an item dependent cluster. We propose a new method to cluster individuals while detecting and correcting for response styles. Specifically, we assume the existence of a response function that relates observed responses based on an individual-specific scale to those based on a common scale and then apply cluster analysis to data transformed based on the estimated function. We use a monotone I-spline function to estimate this response function. The performance of the proposed method is examined by simulation and by empirical data analysis.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2017 program
|
