ABSTRACT.A comparison of one‐way analysis of variance (ANOVA) and cluster analysis as statistical methods for selecting members of a descriptive panel was conducted. Using 150–mm unstructured line scales, 10 panelists evaluated the performance in coffee of 2 whiteners on 14 sensory attributes. Factor analysis showed that 5 factors were adequate to model the data. The sensory ballot used had a reliability coefficient omega (Ω) equal to 0.93. One‐way ANOVA and cluster analysis resulted in 2 different panels consisting of 9 judges each. Means and standard deviations of the scores showed that greater variability among panelists’ ratings resulted when the panel was selected using one‐way ANOVA than when cluster analysi