Susunan Kuadrat Latin Untuk Taste-Panel Experiments

Abstract

A taste-panel experiment is an experiment where several panellists score the taste of food using an ordinalscore, e.g. from 1 (extremely tender) to 8 (extremely tough). To avoid an order effect of the panellistposition, we apply a Latin square arrangement such that each sample appears once in each position. If thenumber of panellists available is less then the number of treatment combinations provided then we have anincomplete block design since we do not have a complete replication. Hence, these arrangements may not bebalanced. We use a special Latin square, the so called Williams design, to avoid a first carry-over effect andorder effect. The design can be balanced using an extension of William's design. Finally, the arrangement oftreatment combinations was randomly permuted to avoid systematic bias due to the generation of Latinsquares.

Key words: carry-over effect, ordinal data, Latin square, outliers, position effect, taste panel experiments