Estimación de las preferencias del consumidor a través de un modelo jerárquico bayesiano logístico multinomial

Abstract

The statistics techniques have won importance in market research, this techniques usually use when the objective wants to know the consumer's factors that a ects decisions of consumer. Many methodologies are used how conjoint analysi, maxdi and trade o . But the simples techniques don't have optimal results and other techniques have a long duration, for this reason this techniques can give a bad results. In that job, we propose a di erent trade o methodology, that methodology compares two supplies and the subject selects one, again the last selection compares with other supply and nally the subject compares all supplies. If the numbers deals are k, the subject have to compare k1 deals, for that reason that methodology can reduce the times and the costs. For estimate the probability that the subjet i(i = 1; :::; n) select the supply j(j = 1; :::; k, we propose a Bayesian hierarchical multinomial logit model using MCMC (Markov Chain Monte Carlo), with that model we can estimate the probability that the subject i select each supply when the subject couldn't compare all supplies. When we are found the probabilities of each subject, we can apply a principal component analysis and in this manner we can do a classi cation of the subjects where nally we can identify markets niches for the di erents supplies.