紀要論文

A Comparative Study of Data Gathering Procedures in Conjoint Measurement

eng

09169180

https://doi.org/10.15002/00020855

法政大学産業情報センター

グノーシス：法政大学産業情報センター紀要＝Γνωσις

2

18

33

1993-03-31

An experiment is designed for testing validity and reliability of two data gathering procedures in conjoint measurement. Computer-interactive Adaptive Conjoint Analysis (ACA in short) and the conventional Full-Profile method (FP in short) are among those compared for predictive performance. After responding to four questionnaires, two data collection procedures each for two product categories (chocolate and soft drink), in a computer-assisted session, two hundred and six respondents picked up their most favorite brand(s) from a set of brands with relatively high shares in the market. For soft drink category, partworths of product attributes are estimated for price, manufacturer, brand category, and size of container. For chocolate, importance weights are estimated for price, maker, taste, and product form. Average correlation coefficients between parameter estimates derived from the different data collection procedures (ACA and FP) are quite high; 0.52 on the average for both product categories, mostly above 0.65 for individual respondents. Using parameter estimates, total utility scores could be calculated for the brands presented at the final stage of computer interview. Then, the first choice could be predicted and matched with the brand actually picked up by each respondent. "Batting Average" for FP method is 53.9%, which is fairly higher than 44.7% for ACA procedure in predicting the choice of chocolate. However, ACA with an average of 45.7% could hit the right cans of Cola, Tea, or Orange Juice better than FP only with an average of 40.4%. We recommend that researchers would better make use of ACA against FP, when there are many attributes and/or profiles, since interviewing with ACA is much easier than that with FP.

Article

AN10380131