Analysing multivariate storage data of seafood spreads. A case study based on combining split-plot design, principal component analysis and partial least squares predictions

Abstract

As part of an extended fish product (mixed salmon spread) value chain involving multiple treatment procedures and mixing processes, oxidation and microbial spoilage can be initiated at any number of steps and go on to accelerate product deterioration. This may occur, for example, when salmon rest raw materials are processed to form mixed emulsion products. To investigate the effect of selected variables in the value chain, a model experiment was designed and implemented, consisting of a chain divided into four steps involving fish feed composition, fish processing, fish spread production and storage. By using this case, the objectives of the paper are to 1) show how a complex split-plot design can be analysed using analysis of variance (ANOVA) and multivariate statistical analyses, 2) show how an interplay of the methodologies can contribute to improvement in the interpretation and validation of results, and 3) identify the quality markers most affected by the design variables, and then use these to optimise response measurements for different raw material properties. We also propose some new monitoring and control strategies based on the PCA and results obtained. The analysis has indicated in this case that it may be beneficial for the long shelf-life of the spread to use fresh and lean salmon cuts, to store the product under superchilled conditions and to avoid the addition of secondary seafood ingredients. Salmon feed variables do not affect the eating quality of the spreads. The early addition of a smoke component and the rigor status of the salmon at the time of processing had little effect on eating quality. The variables that did not affect eating quality or shelf-life can be optimised based on aspects such as nutritional or health benefits, or production costs. This article demonstrates that PCA is a useful method both for the monitoring of eating quality with storage time, the definition of control limits for product acceptability, and the statistical validation of split-plot ANOVA results. • Identification of significant variables along the mixed salmon spread value chain. • Inability to predict eating quality using measurements taken early in the value chain. • PCA identified as a useful tool for monitoring eating quality development. • PCA identified as a useful tool for validating split-plot ANOVA results.