Young-Seung Lee, Jean-François Meullenet
Abstract
This study was conducted to investigate the impacts of eliminating first order samples and data corrections on overall liking (OL) results. Nine consumer studies were employed for this work. For each study, all first order products were eliminated from the data to examine the impacts of first order bias on OL. Significantly higher OL values were observed for products presented in first order for 6 of the 9 studies. After removing first order samples, OL for each product slightly decreased. In addition, slight changes in sample OL rank orders were observed for 8 of the 9 studies. Data sets with first position sample removed seemed less discriminative. This indicates that samples were discriminated to a greater extent due to first order bias and actual differences might be overstated. Two first order data corrections were successfully adapted to remedy the first order bias even after data were segmented by cluster analysis.
source: visit here
