Imputation Procedures

One of the features of NCC Food and Nutrient Database that makes it the choice of researchers is the small number of missing nutrient values for foods in the database. To minimize the number of missing values, NCC uses several standardized procedures to impute or logically calculate estimations. These procedures are described in detail in the article Procedures for Estimating Nutrient Values for Food Composition Databases. To summarize, one of the following procedures is employed to estimate a nutrient value for a food when a value is not available from the USDA National Nutrient Database for Standard Reference (SR) or cannot be found in the literature:

  • Use value from a different but similar food (e.g., missing nutrient for wild duck may be judged to be the same as known nutrient in domestic duck)
  • Calculate value for another form of the same food (e.g., convert from raw to cooked values using retention factors)
  • Calculate values from other components in the same food (e.g., estimate total vitamin A from provitamin A, carotenoids, and retinol)
  • Calculate value from household recipes or commercial food product formulations for multi-component foods. This process involves building foods from basic food items (Core Foods). This is a relatively simple process for foods derived from household recipes (e.g., homemade lasagna) because both the ingredients and ingredient amounts are specified in recipes. The process is more complex for brand name food products for which ingredient information is generally known from the ingredient statement on product packaging, but the amount of each ingredient is not specified. Consequently, a special procedure is used for calculating nutrient values for brand name food products.

 

Missing values are allowed in the database for foods that are consumed in very small quantities, such as spices, or where there are no data to indicate whether the nutrient exists in the food.

 

Some nutrients have no missing values, but a high percentage of imputed values. An example is total vitamin A which is calculated from provitamin A, carotenoids and retinol.

 

In future versions of the database, missing and estimated values will be replaced by analytic values as they become available.