Dietary Pattern Analysis

Dietary pattern analysis is the nutritional-epidemiology application of multivariate statistics that identifies how foods are actually eaten together, summarizing the whole diet into a few empirical patterns rather than studying single nutrients in isolation. Introduced as a research direction by Frank Hu in his 2002 Current Opinion in Lipidology review and surveyed methodologically by Newby and Tucker in 2004, the approach takes a matrix of food-group intakes and applies factor (principal component) analysis, cluster analysis, or reduced-rank regression to extract a posteriori patterns such as a 'prudent' pattern rich in fruits, vegetables, and whole grains and a 'Western' pattern high in red meat and refined foods. While the underlying algebra is generic principal component or cluster analysis, what makes this a distinct method is its substantive construction: the input is the food-group intake matrix of the whole diet, and the output is interpretable eating patterns linked to disease.