Design of sensory multi-session trials with preparation constraints
Design of sensory multi-session trials with preparation constraints Designs for sensory studies must satisfy several requirements. Usually a given number of products are to be evaluated and there is an upper limit to the number of assessors available. Due to variation in sensory perception, inter-assessor product comparisons are preferred. For large product numbers, trials are split into sessions to avoid sensory fatigue and the sequential presentation of products can cause order and carry-over effects. Thus, resolvable row-column or cross-over designs are required, which ensure that each assessor tastes all products the same number of times. In this thesis a three-step procedure is proposed to generate designs for trials where the number of products prepared for or served in each session is limited. First, an incomplete block design with a special column structure, the preparation design, is created, assigning products to sessions. Secondly, a cross-over design is constructed, assigning the columns of the preparation design to assessors. In the third step the two designs are combined by identifying the column-order of the preparation design that results in the highest average efficiency of the complete cross-over design. Search algorithms for incomplete block and cross-over designs are modified to produce preparation and panel designs with a special structure to guarantee resolvability of the complete sensory design. This procedure has been enhanced to produce designs for trials involving a control and several test products, in which control-test comparisons are estimated with higher precision than test-test comparisons. Two distinct construction methods have been developed for this case. By using factorial preparation designs the three step procedure can also be adapted for creating factorial multi-session designs with or without a control product.