Statistical validation and criterion for paired comparison data in sound quality evaluation

In statistical analysis of paired comparison data in sound quality evaluations, the commonly adopted Kendall's coefficient of consistency criteria is not sufficient for assessment of judging errors. Based on the analysis of misjudging rates in the same-sound-event-comparison, the different-replay-order-comparison and the circular triads, an accurate calculation model for circular triads is proposed. For validation of jurors' test data, these three types of errors are considered, and a criterion called weighted factor of consistency is developed. The proposed methods and criterion have been applied to the analysis of subjective data of the paired comparison for car interior sound quality evaluation. Good agreements between analyzed results and Bradley-Terry model fitting results are acquired, and rightness of proposed criterion has been proved.