Food extrapolation

If the food extrapolation setting has been checked, extrapolation of concentrations is performed for all food/active substance combinations for which:

  1. the number of measurements in the analytical scope is smaller than a given threshold for extrapolation (default 10), and

  2. there is an extrapolation rule allowing extrapolation of concentrations from one or more other foods (the from-food(s)) to the given food (the to-food), and

  3. (optional criterion:) the substance is associated with authorised use for both foods, and

  4. (optional criterion:) concentration limits (e.g. MRLs) on the from-food and to-food exist and are equal. Note: if the active substance is not a measured substance, then the MRL check has to be made per measurement at the level of the measured substance which provided the concentrations assigned to the active substance.

Food extrapolation is performed by one of the following procedures: 1) Substance-specific imputation of missing values by extrapolated measurements, or 2) Extrapolation of complete samples for multiple substances.

1. Substance-specific imputation of missing values by extrapolated measurements

The missing values in the active substance concentrations of the to-food are imputed in a random order by active substance concentrations (positive, censored or zero) from a randomised list obtained from the fromfood(s). By matching the randomised lists, each from-food measurement is assigned at most once, so after extrapolation there may still be missing values left, or not all measurements of the from-food(s) may have been used for extrapolation.

Note: In this method, it is assumed that the to-food has a sufficient number of samples. No extrapolation is applied for foods with no samples at all, and data gaps will also remain for foods with fewer than n samples, because no new samples are added.

Note: the resulting occurrence patterns will be random with respect to the extrapolated substances, i.e., observed occurrence patterns for the from-food are not extrapolated to the to-food.

2. Extrapolation of complete samples for multiple substances

(not yet implemented)

All samples of the from-food(s), i.e., complete samples with data for all active substances, are copied as samples for the to-food and added to the existing to-food samples. For example, extrapolate all apple sample records to the available pear sample records. However, measurements for substances that do not fulfil the (optional) criteria 3 and 4 above are non-valid extrapolations and are replaced by missing values. The status of the extrapolated samples is stored to distinguish between extrapolated and non-extrapolated sample records. Note that this method maintains correlations in the occurrence patterns and postpones imputation of MVs until the concentration models step.