Resumen de Beverage Complexity Yields Unpredicted Power Results for Seven Discrimination Test Methods
David J. Bloom, Hwa-Young Baik, Soo-Yeun Lee
The power of discrimination tests is crucial in determining sample size and resources needed for testing. Although research has been conducted on the power analysis of several discrimination testing methods, much of the previous research has focused on basic taste solutions, which may not be directly applicable to food and beverage systems. The objective of the current study was to compare the power of seven discrimination tests: Panelist-Articulated-2-Alternative Forced Choice (PA-2-AFC), triangle, triangle with partial presentation, duo-trio, duo-trio with partial presentation, 4-category rating methods for R-index measure, and same-different pairwise comparison for R-index measure using commercial-type beverage products. Sixty-one prescreened panelists participated in the study. Six product comparisons were performed using tea, tomato juice (three comparisons), citrus-flavored carbonated soda, and cola-flavored carbonated soda. The tests were randomized over two testing sessions for each product comparison. Triangle testing methodologies were found to be overall the most powerful methods across product categories. The PA-2-AFC method was found to be the least powerful across all products. Thurstonian modeling predicts that the PA-2-AFC method would be the most powerful method contrary to the findings of the current study. The products tested were complex in both basic formulations and in changes made between control and variant samples. Complexity of the products may have influenced the discriminability by the panelists using different discrimination tests. Further research should be conducted to characterize the specific influence of sample complexity on the power of discrimination methodology. Practical Application There are several discrimination testing methods that can be selected when determining whether two products are significantly different. A method with high statistical power can allow researchers to save time and resources when addressing this question. The current research compares seven discrimination test methods in order to determine which method results in the highest power for several common commercial-type beverage products. The results from this study demonstrate deviations from Thurstonian model predictions of method power revealing the need to experiment with several methods using commercial-type products commonly tested within a business or research setting prior to selecting an optimal method to use.
