Materials and Methods (MM) Section 1: Linear Regression Model

The intake rates (R_i) in units of mg/kg bodyweight (BW)/day for chemical i were approximated as:

where a₀ represents a “grand mean” intake rate over all chemicals that is unexplained by the exposure predictors, δ_ij is a Boolean (0/1) value that represents whether or not exposure to chemical i occurs via pathway j, a_j represents the additional mean intake rate via pathway j over all chemicals with exposure via pathway j that is unexplained by the predictors, and w_j,k represents the loading (“weight”) of predictor k for chemical i by pathway j (π_i,k). w_j,k is zero if an exposure predictor does not correspond to pathway j. The exposure predictors and intake rates inferred from NHANES data were scaled as X′=X−X¯sd(X) so that their value indicates the number of standard deviations (sd(X)) above or below the average (X¯) for predictor or rates X. This scaling makes the w_j,k directly comparable to each other.