The proposed research will 1) document the accuracy of a customary technique for estimating the levels of nutrients in composite foods, which is trial-and-error estimation of ingredient proportions; and 2) develop and validate mathematical techniques for performing this type of estimation of nutrient levels. These activities relate to eligibility criteria #2 and #5 for the Small Grants Program in Epidemiology. The trial-and-error technique is commonly used by nutritionists to estimate nutrient values of composite foods. Two mathematical techniques for performing such estimations will be applied and validated: linear programming and quadratic programming. The hypotheses to be tested are that the mathematical techniques are significantly more accurate and productive in performing nutrient value estimation than is the trial- and-error technique, and that quadratic programming is more accurate than linear programming. The accuracy of the three techniques will be evaluated by comparing the estimated nutrient values to nutrient values obtained via chemical analysis. The productivity level of nutritionists using each of the techniques will also be compared. The proposed research has potential for improving the quality of cancer epidemiologic research by improving the accuracy of food nutrient databases. Such an improvement would allow epidemiologic researchers to better investigate the possible links between dietary intake and various types of cancer.
