Bayesian Modelling in Conjoint Analysis

Bajesijansko modeliranje u analizi združenih efekata

Vesna Janković-Milić,

Vinko Lepojević

📄 Download PDF

Issue:

41/3

DOI number:

UDK 519.23:658.8,

Abstract

Statistical analysis in marketing is largely influenced by the availability of various types of data. There is sudden increase in the number and types of information available to market researchers in the last decade. In such conditions, traditional statistical methods have limited ability to solve problems related to the expression of market uncertainty. The aim of this paper is to highlight the advantages of bayesian inference, as an alternative approach to classical inference. Multivariate statistic methods offer extremely powerful tools to achieve many goals of marketing research. One of these methods is the conjoint analysis, which provides a quantitative measure of the relative importance of product or service attributes in relation to the other attribute. The application of this method involves interviewing consumers, where they express their preferences, and statistical analysis provides numerical indicators of each attribute utility. One of the main objections to the method of discrete choice in the conjoint analysis is to use this method to estimate the utility only at the aggregate level and by expressing the average utility for all respondents in the survey. Application of hierarchical Bayesian models enables capturing of individual utility ratings for each attribute level.

Apstrakt

Statistička analiza u marketingu u velikoj meri je uslovljena dostupnošću različitih tipova podataka. Poslednjih decenija zabeležen je nagli porast u broju i vrstama podataka koji su dostupni istraživačima tržišta. U takvim uslovima tradicionalni statistički metodi imaju ograničene mogućnost za rešavanje problema vezanih za izražavanje neizvesnosti na tržištu. Cilj ovog rada je da ukaže na prednosti primene bajesijanskog zaključivanja, kao alternativnog pristupa klasičnom zaključivanju. Metodi multivarijacione statistike pružaju moćne alate za postizanje brojnih ciljeva marketinških istraživanja. Jedan od tih metoda je i analiza združenih efekata, koja obezbeđuje kvantitavnu meru relativnog značaja jednog atributa proizvoda ili usluge u odnosu na drugi atribut. Primena ovog metoda podrazumeva prikupljanje informacija o potrošačima, pri čemu oni izražavaju svoje preferencije, a statistička analiza obezbeđuje numeričke pokazatelje korisnosti svakog atributa. Jedna od glavnih primedbi na metod diskretnog izbora u analizi združenih efekata je da primena ovog metoda omogućava procenu korisnosti samo na agregatnom nivou i to izražavanjem proseka korisnosti za sve ispitanike u istraživanju. Primena hijerarhijskih bajesijanskih modela omogućava obuhvatanje individualnih ocena korisnosti za svaki nivo atributa.

References