Issues in Assigning “No Preference/Difference” Responses

Andy Basehoar
Senior Statistician, Family Care

Analysis of preference questions has traditionally utilized a binomial model to assess statistical significance thresholds for claims substantiation. As the ASTM standards have evolved, now incorporating splitting the “no preference” response for superiority claims, there is an opportunity explore other statistical approaches. The current and alternative statistical approaches to handling “no preference” data and their advantages and disadvantages are discussed. These methods may be useful to resolve some of the paradoxical results from the E1958 superiority and unsurpassed standards.