Processing factor correction

Processing factors can be specified as fixed factors (nominal) or as statistical distributions for the variability across samples.

Concentrations in the consumed food (food as eaten) may be different from concentrations in the food as measured in monitoring programs (typically raw food) due to processing, such as peeling, washing, cooking etc. Concentrations are therefore corrected according to

\[c'_{jhk} = \mathit{pf}_{jhk} \cdot c_{jhk} = \left( \frac{\mathit{PF}_{k}}{cf_{k}} \right) \cdot c_{jhk}\]

where \(c_{jhk}\) is the concentration of substance \(k\) in the food \(j\) with processing type \(h\), and where \(\mathit{pf}_{jhk}= \frac {\mathit{PF}_{jhk}}{cf_{jhk}}\) is a factor indicating the mass change for a specific combination \(k\) of food as measured and processing. The processing correction factor \(\mathit{cf}_{jhk}\) is used to correct for the fact that the processing factors \(\mathit{PF}_{jhk}\) as commonly available from the input data describe both the effects of chemical alteration and weight change. E.g. for a dried food with a consumption of 100 gram which is translated to 300 gram raw agricultural commodity, the correction factor is 3. Note that the weight change is already included when calculating the consumption amounts of the foods-as-measured.

  • The distribution is either the logistic-normal distribution for processing types with factors restricted between 0 and 1 (e.g. washing),

  • or the lognormal distribution for processing types with non-negative factors (e.g. drying).

Variability distribution parameters are specified indirectly via the 50th and 95th percentile. Uncertainty for processing factors can be specified using uncertainty distributions of the same form as for variability. Uncertainty distribution parameters are specified indirectly via the 95th uncertainty percentiles on the 50th and 95th variability distribution percentiles.

For distribution based processing factors specify \(f_{k,nominal}\) and \(f_{k,upper}\) (Nominal and Upper in table ProcessingFactors). Two situations are distinguished depending on the type of transformation.

Nonnegative processing factors

Equate the logarithms of \(f_{k,nominal}\) and \(f_{k,upper}\) to the mean and the 95% one-sided upper confidence limit of a normal distribution. This normal distribution is specified by a mean

\[ln(f_{k,nominal})\]

and a standard deviation

\[ln(f_{k,upper}) – ln(f_{k,nominal})/1.645\]

Processing factors between 0 and 1

Equate the logits of \(f_{k,nominal}\) and \(f_{k,upper}\) to the mean and the 95% one-sided upper confidence limit of a normal distribution. This normal distribution is specified by a mean

\[logit(f_{k,nominal})\]

and a standard deviation

\[logit(f_{k,upper}) – logit(f_{k,nominal})/1.645.\]