SENSORY THRESHOLDS: ESTIMATION FROM DOSE-RESPONSE CURVES

Three methods of threshold estimation are compared for the detection of 1,8-cineole added to Concord grape juice presented in a semi-ascending paired difference design (Lundahl et al. 1986). Method A tests the significance of obtaining a nonzero response, method B determines the transition point of a 2-phase linear model fitted to the data, method C fits a logistic model to the data and defines threshold as the point of maximum curvature, C_max, of the curve. Method A yields interval estimates of thresholds, the midpoints of which systematically decrease with increasing sample size. Methods B and C yield point estimates of thresholds which do not systematically decrease with increasing sample size, are a feature of the dose-response curve, and are similar in value to each other.