RT Generic
T1 Reconsidering optimal experimental design for conjoint analysis
A1 Esteban-Bravo, Mercedes
A1 Leszkiewicz, Agata
A1 Vidal-Sanz, Jose M.
A2 Universidad Carlos III de Madrid. Departamento de Economía de la Empresa, 
AB The quality of Conjoint Analysis estimations heavily depends on the alternatives presented in the experiment. An efficient selection of the experiment design matrix allows more information to be elicited about consumer preferences from a small number of questions, thus reducing experimentalcost and respondent's fatigue. The statistical literature considers optimal design algorithms (Kiefer,1959), and typically selects the same combination of stimuli more than once. However in thecontext of conjoint analysis, replications do not make sense for individual respondents. In thispaper we present a general approach to compute optimal designs for conjoint experiments in avariety of scenarios and methodologies: continuous, discrete and mixed attributes types, customerpanels with random effects, and quantile regression models. We do not compute good designs, butthe best ones according to the size (determinant or trace) of the information matrix of theassociated estimators without repeating profiles as in Kiefer's methodology. We handle efficientoptimization algorithms to achieve our goal, avoiding the use of widespread ad-hoc intuitive rules.
YR 2012
FD 2012-11
LK https://hdl.handle.net/10016/14548
UL https://hdl.handle.net/10016/14548
LA eng
NO Research funded by two research projects, S-0505/TIC-0230 by the Comunidad de Madrid andECO2011-30198 by MICINN agency of Spanish Government
DS e-Archivo
RD 17 jul. 2024