A SIMPLE ALTERNATIVE ANALYSIS FOR THRESHOLD DATA DETERMINED BY ASCENDING FORCED-CHOICE METHODS OF LIMITS

An alternative analysis of forced-choice threshold data sets such as the type generated by ASTM method E-679 involves a simple interpolation of chance-corrected 50% detection. This analysis has several potential advantages. The analysis does not require the ad hoc heuristics for estimating individual thresholds above and below the series. It takes into account the possibility of guessing correctly, which is not considered in the ASTM calculations and produces a downward bias to the estimates. It does not discount correct responses early in the series which may be legitimate detections, but which are discounted by the ASTM method if followed by any incorrect response. Comparisons of the two methods in a large consumer study of odor detection threshold study data set gave comparable values. The interpolation can also be done to determine other levels of detection (e.g. 10, 25%). These values other than 50% can be potentially useful in setting regulatory standards for water or air pollution limits or to food manufacturers who wish to avoid detection of taints by more sensitive individuals. PRACTICAL APPLICATIONS The forced choice methods for threshold estimation have proven practically useful in comparing the potency of various flavor materials and in comparing the sensitivities of individuals. The ASTM method E-679 is one such method. The alternative analysis of results from this procedure which is outlined here provides additional information and does not exhibit the downward bias because of correct guessing.