Number of judges necessary for descriptive sensory tests

• Ideal number of judges for descriptive evaluations was calculated by “power analysis and sample size” procedure.
• Previous researches in the sensory descriptions were utilized.
• The RMSE from 574 previous researches were collected.
• Three different levels of type I and type II errors were considered.
• Were simulated 180 scenarios of descriptive tests.

To determine the number of judges needed in descriptive tests four parameters are necessary: the probability of type I error (α), the probability of type II error (β), the difference in averages which is sought in the experiment (d’) and standard deviation of the experimental error (s). Probabilities of the experimental errors and the difference that one desires to detect between means should be stipulated by the researcher. Therefore, only the estimate of the experimental error cannot be previously obtained, which must be obtained experimentally or by means of similar researches previously performed. Because in descriptive sensory analysis the most common approach to data analysis is the analysis of variance, the estimated standard deviation of the experimental error is obtained by the root mean square error (RMSE). Therefore, 574 RMSE values were obtained from previous published studies. The data collected was adjusted to Weibull probability distribution (1.8081, 0.11419), where five percentiles of the distribution are considered in the calculations. Determination of the number of evaluations necessary was performed using the procedure “sample size and power analysis” of the JMP/SAS software. Three probability levels were defined for type I and II errors, four levels of mean difference to be detected in the experiment, and five percentiles of RMSE distribution probabilities. The required numbers of evaluations in descriptive tests, considering these different experimental conditions, were calculated totaling 180 scenarios. Considering the median of the experimental error, a value of alpha (type I error) of 1%, a value of beta (type II error) of 5% and a difference between the average of 10% of the sensory scale, 33 evaluations are needed in the descriptive tests. Further considering that each judge evaluates the samples triplicate, 11 judges are necessary for this specific set of parameters. Other scenarios were also discussed in the paper.