Common problems in the experimental design of sensory tests in agricultural studies and recommended solutions

Citation

Bejaei, M. (October 2020). Common problems in the experimental design of sensory tests in agricultural studies and recommended solutions. Sensometrics 2020, Stavanger, Norway. (Virtual)

Résumé en langage clair

This presentation was prepared to share specific experimental design considerations required when conducting sensory tests using raw agricultural products. Most of the raw agricultural products are perishable foods, produced under natural conditions, and researchers have little or no control over many production factors that may influence the quality of the products or their sensory attributes. Finding exact replicates in raw agricultural products is not feasible in most cases because between cultivar and within cultivar variations exist among them. Variations even exists within one fruit/vegetable. Greater variation sources between and within experimental units require detailed sampling protocols, a larger sample size, repeated tastings, and validation of the results. Familiarity with studied products, their production practices, postharvest physiology, unique product characteristics and their market are also beneficial in selecting proper experimental designs.

Résumé

Sensory research projects relevant to raw agricultural products (especially fruits and vegetables) can be designed using various research methods. There are many factors, some unique to the products, that should be considered in the design of those experiments. Most of the raw agricultural products are perishable foods, produced under natural conditions, and researchers have little or no control over many production factors that may influence the quality of the products or their sensory attributes. Finding exact replicates in raw agricultural products is not feasible in most cases because between cultivar and within cultivar variations exist among them. Variations even exists within one fruit/vegetable; an attribute can differ from top-to-bottom and side-to-side of the fruit (e.g., sun-exposed side or shaded side). In some agricultural studies, samples were prepared without considering these variation sources. For example, if slices of fruits were provided to judges in a sensory test, those slices were not excised from specific parts of the fruits (e.g., the sun/shade transition zones in apples) to reduce the extra variation sources. In addition, the variations within experimental units required provision of multiple biological replicates from the same treatments for each judge (to capture random variation and evaluate the judge performance), and repetition of the study to collect an independent validation data set to be able to generalize the results.
In some studies, the applied tests did not match the scales of the measurements as a result statistical test assumptions were not met. One of the common observed problems in agricultural studies has been related to the design of tests in which sensory ratings were conducted by the same judges multiple times in an experiment (e.g., in a storage study) nonetheless the data were analyzed without the use of the repeated measures models. For example, they analyzed treatment and time effects using independent tests or in a two-way ANOVA without meeting the independent-measurements assumption of parametric tests. Identifying whether time was a fixed effect or a random effect has not been done in a clear and consistent way in some cases. There were also studies in which a treatment effect was considered as a between-subject factor even though all judges tasted samples from all treatments.
Several examples will be presented to demonstrate the impact of the common research design and statistical analysis problems in studying the sensory characteristics of raw agricultural products and solutions will be discussed.

Date de publication

2020-10-06

Profils d'auteurs