Recipe/Theory/Errors

From Open Food Facts wiki
Jump to navigation Jump to search

When using the Ciqual data, we are using the average(?) values for ingredients. We do not know the actual variability of an ingredient. We need other sources for this (OFF category data?). But we have this how could we use it?

Assume that each nutrient j of each ingredient i has an error σij. The combined error for the nutrient j is then:

(1) σj2 = Σi=0N-1βi2σij2

With:

  • βi the contribution of ingredient i to the recipe;
  • N the number of ingredients in the recipe;

This approach assumes that the ingredients are independent. And thus we have an estimated error for each estimated nutrient. And that the variability is normally distributed. We assume here that the contribution of the ingredient has no error.

What would be the errors on the estimated βi. Or are we then arguing in circles?

References