Table 2.

A contingency table of the proportions of selection by 79 participants for individual sensory attribute terms among coffee samples evaluated at the three different temperatures.

Terms1	Sample temperatures			Q-value	P-value	Cramér’s V value
	5°C	25°C	65°C
Aroma
Pungent	0.19a	0.06b	0.14ab	6.08	0.049	0.15
Roasted	0.35b	0.52ab	0.54a	6.86	0.03	0.17
Taste/flavor
Metallic	0.33a	0.22ab	0.15b	7.74	0.02	0.17
Roasted	0.33b	0.37b	0.63a	18.00	<0.001	0.27
Skunky	0.19a	0.06b	0.08b	9.58	0.01	0.18
Bitter taste	0.84a	0.84a	0.67b	8.05	0.02	0.19

Cochran’s Q-test (Cochran, 1950), using the exact probability and distribution of the Q statistic (Patil, 1975), was performed to determine whether the proportions of participant selection for individual terms of sensory check-all-that-apply (CATA) question could differ by sample temperature. The proportions with different letters within each row represent a significant difference determined by post hoc multiple pairwise comparisons using the Marascuilo procedure (Marascuilo and McSweeney, 1967). Cramér’s V value was used to measure strength of association between two nominal variables (sample temperature × selected/unselected case) for the contingency table. Cramér’s V values, ranging from 0 (no association between the variables) to 1 (perfect association), of 0.1, 0.3, and 0.5 were considered small, medium, and large associations, respectively (Cohen, 1988). ¹Only significant terms, determined by Cochran’s Q-test, among 49 sensory attribute terms were shown (P < 0.05).