Conjoint Analysis: How many attributes should be used in a paired comparison task?
An empirical examination using a new validation approach.

Abstract:

Conjoint analysis is one of the mostly used methods in the analysis of preferences and the prediction of choices. Today, most applications of conjoint analysis use paired comparisons. Paired comparisons on a graded scale are a substantive element of the adaptive conjoint analysis (ACA), a computer supported procedure that is predominantly used in marketing research. Contrasting this heavy application, little is known about how many attributes to present in a pair and how many judgments to be made.
We present an empirical study testing the validity of paired comparison tasks using two, three, and four attributes. This examination has been made by using a new approach of validation. Our results show that paired comparisons with two attributes lead to more accurate estimations than with three or four attributes. Furthermore, pairs with two attributes demand less time and are subjectively considered more beneficial in some sense.

Download:

Conjoint Analysis: How many attributes should be used in a paired comparison task?
An empirical examination using a new validation approach.

Kontakt Reiners >>

Weitere Publikationen zur Conjoint Analyse von Wolfram Reiners:

1999. Evaluation verschiedener Formen von Paarvergleichen in der Conjoint Analyse.

1989. How many attributes should be used in a paired comparison task? An empirical examination using a new validation approach.

1996. Multiattributive Präferenzstrukturmodellierung durch die Conjoint Analyse. Diskussion der Verfahrensmöglichkeiten und Optimierung von Paarvergleichsaufgaben bei der adaptiven Conjoint Analyse.