Factor Analysis as a Technique for Examining Complex Data on Foodstuffs

Abstract

In a recent article in this journal I discussed certain of the fundamental problems involved in the subjective appraisal of foodstuffs.4 The observations made were not primarily concerned with any one particular food product, although interest in the whole set of problems arose out of a number of years' work on the skills of professional cheesemakers and cheese graders. The question was raised as to how one should examine a set of 'complex multi-dimensional relationships' such as may exist between 'objective' test data and skilled judgements on a number of cheeses. In the particular investigations to be considered the objective tests were primarily mechanical in nature, but such data might equally well have included information derived from chemical or bacteriological tests. In the previous paper it was suggested that one method of dealing with the particular type of problem which has just been posed is the technique of Factor Analysis. A brief outline will be given of the main features of this technique, and its use with food products will be illustrated by these studies of cheeses. It is of interest to add that very recently Baker' has used factor analysis to examine the connection between data derived from various forms of routine chemical analysis and data derived from the evaluation of the quality of samples of wine by normal sensory methods. This particular study involved less than twenty samples of wine, whereas almost two hundred cheeses were used in the studies which will be summarised and discussed later. The main function of the present article is to examine both the potentialities and the limitations of factor analysis as applied to data concerning foodstuffs. No systematic examination along these lines has yet been made. Ostle and Tischer'0 in their otherwise very comprehensive review of statistical methods in food research make no reference whatever to factor analysis. Yet it is seven years since the first preliminary study employing data on cheeses was published (Harper and