Aroma extract dilution analysis. Precision and optimal experimental design

The odor thresholds of 12 different compounds have been determined in an AEDA experiment using a panel composed of 8 individuals. Only in one case is the distribution of thresholds among judges significantly different from the log-normal. The cause of that departure from normality seems to be a cross adaptation rather than anosmia. The standard deviations (SD) range from 2(0.7) to 2(4.1), with 2(1.8) as average. If the AEDA is carried out at a dilution rate, R, and dilution R(P) (where P = 0, 1, 2.) is the last one at which a compound was detected by a judge, the flavor dilution (FD) factor that should be provided for that judge is R(P+0.5). In the case where several judges carry out the AEDA, the FD for a given compound should be the geometric mean of the FD provided by all the judges. The SD of the distribution of FDs is related to that of the distribution of odor thresholds, but is strongly influenced by the dilution rate, R, being higher with higher R values. The relationship between both SDs can be used to determine the expected precision (measured as the geometric length of the 95% confidence interval) of the result of an AEDA experiment as a function of the dilution rate, the number of judges, and the SD of the distribution of thresholds. Different simulations have shown that in most cases, a dilution rate of 10 is the best option, and that lower dilution rates are advantageous only if the analyzed solution contains compounds with a very narrow distribution of thresholds.