Designing your sensory test for simple and effective statistical analysis

Chantal C. Gilbert
Campden & Chorleywood Food Research Association, United Kingdom
 c.gilbert@campden.co.uk

 

Experimental design is an important component of any scientific research including sensory evaluation. The data collected in sensory and consumer sciences are obtained from responses to stimuli as perceived through the senses. The assessors who generate these data can be a significant source of variation, which needs to be accounted for in the analysis. Furthermore, because of the nature of this data, experimental design takes on added importance because of the need to ensure that other extraneous sources of variation are randomised or controlled for during experiments.

Experimental design is more than just sample presentation order, and includes:

• Specifying of experimental and statistical objectives
• Selection of sensory test method
• Choice of scales
• Practical constraints, such as the need to cook and serve samples warm
• Identification of factors of interest (for example, storage conditions, packaging types, ingredients types and levels)
• Choice of sample presentation design (such as randomised complete block designs, completely randomised designs or balanced incomplete block designs)
• Planning for correlation of sensory results with other information, such as consumer or chemistry data.

Decisions made at each of these stages are directly linked to the choice of statistical analysis. For example, the choice of statistical model used in an Analysis of Variance is directly linked to the sample presentation design used to collect the data.

An example will be shown where two different types of gelatin substitutes were selected and compared with gelatin. A full factorial design was used, combining type of stabiliser (gelatin, replacement A, replacement B), concentration (high, medium, low) and sugar (high, low), creating 18 samples of mousses. Descriptive profiling of the mousses was carried out using a trained sensory panel. In addition, the rheology of the mousses was measured and correlated with the sensory texture attributes. The choice of a full factorial design provided an efficient means of analysis, where the ingredients and concentration levels could be individually evaluated and the interaction between them assessed.

 Using this and other examples from sensory and consumer research, the link between design and statistical analysis will be demonstrated.

Download presentation