| Food Composition Data: A User's Perspective (1987) |
|The uses of food composition data|
|Using food composition data to communicate nutrition to the consumer|
Using NUTREDFO for nutrition guidance research
Using NUTREDFO food composition data we applied mathematical clustering algorithms to the classification of foods within commodity groups into subgroups or clusters with similar nutrient compositions . The results obtained depend on the nutrients selected. By using nutrients that have limited availability in the food supply (i.e. vitamin B6, calcium, iron, magnesium, folacin, and zinc) and those that pose a possible increased health risk (i.e. sugar, fat, cholesterol, and sodium), it is possible to identify quickly those foods that provide adequate amounts of essential nutrients and excessive amounts of nutrients of concern. Furthermore, the clustering algorithm overcomes a problem that has made it difficult in the past to group foods objectively and accurately, namely, that of dealing simultaneously with more than one or two nutrients. As many nutrient attributes as desired can be used and analysed simultaneously by the algorithm.
The algorithm also provides a "cluster centre" or prototype nutrient composition which represents a summary of the nutrients in the foods assigned to a cluster. Figure 1 illustrates clusters for dairy foods: low-fat milks, plain yoghurt, and buttermilk clustered together owing to their high nutrient density and lowest amounts of total and saturated fat and sugar, and low cholesterol contents (fig. la); whole milk and natural cheeses grouped together due to their moderate nutrient levels and relatively high fat and sodium contents (fig. 1b); both creamed and low-fat cottage cheese clustered together (fig. 1d), with their high sodium content overriding the differences in fat content of these two products. Subgroups based upon similarities in attributes were also identified in other food commodity groups .
The results obtained indicate that this technique provides valuable insight into the nutrient composition of the food supply. Many of the clusters obtained were ones that might have been anticipated. However, some unexpected associations occurred, which, when seen, were quite logical, but would probably not have been predicted. Moreover, in some commodity groups the cluster centres indicate that although the amounts of certain nutrients may vary from one cluster to another they tend to occur in the same proportion. This means that further investigation could lead to a system of "leader" nutrients, those whose presence indicates the presence or absence of other nutrients. The development of the expanded NUTREDFO data base was and will continue to be critical to furthering our research in the area of dietary guidance and nutrition education.
Nutrient density characterization of five dairy group clusters. Food items listed are those