Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Designing conjoint choice experiments using confounded factorial designs
Conjoint choice experiments help researchers understand how people make complex judgments such as purchase decisions and product valuation by posing a series of choices about products or services. In conjoint choice experiments, a respondent is asked to evaluate a series of choice sets. The respondent's task is to choose the most preferred alternative from a series of choice sets. The data generated from conjoint choice experiments can be used to predict marketplace behavior and design products that maximize consumer acceptance. ^ Conjoint choice designs thus far have been based on single fractions of full factorial designs that generally do not allow estimation of interaction effects and produce biased estimates of main effects when interactions are not negligible. As an alternative to fractional factorials, we considered confounded factorial designs that arrange a complete factorial experiment into blocks, where the block size is smaller than the total number of treatment combinations. Confounded factorial designs are proposed for 2 n, 3n, 2n (2α)m, 3n(3β) m, 2n3m, and 2n3m4 o factorials for conjoint choice experiments which allow estimation of main effects and first order interactions. Bias of main effect estimates of some previously proposed fractional designs is found to be large if interactions effects are large relative to the main effects. ^ The log-odds transformation method and the generalized linear mixed model approach have been suggested for the analysis of confounded factorial conjoint choice experiments. The log-odds transformation method is capable of estimating complex covariance structures relevant to confounded factorial conjoint choice experiments, but large sample sizes are required. Although the generalized linear mixed model explicitly utilizes the multinomial distribution, it often will not converge with the complicated covariance structure often implied by using confounded factorial conjoint choice experiments and thus, simpler covariance structure must be used. In the analysis of the beef steak experiment, the log-odds transformation method found more significant effects than the generalized linear mixed model. However it is not known if these effects are due over rejection or due to higher power of the test. Further investigation is needed to evaluate the statistical characteristic of the log-odds model compared to the generalized linear mixed model approach. ^
Yong, Chin Khian, "Designing conjoint choice experiments using confounded factorial designs" (2004). ETD collection for University of Nebraska - Lincoln. AAI3152623.