We want to rank the preference of population of animals for n different recipes. Recipes are tested in two-choice assays, where groups of over 100 individuals can choose one choice or the other. Preference for the group is computed as a preference index, a continuous variable that is between -1 and 1. Is there an experimental design that would help reduce the number of combinations to test in order to rank the recipes ? This seems to me similar to a choice design with a number of survey, the fundamental difference being that the response is not binary, but continuous.
In this experimental approach, it is not possible to test the subjects individually, and we don't have access to individual choices. A non-choice design will be used as a screening method, but non-choice behavioural response is not informative enough.
Thanks for your suggestions!
