Substance contributions to components

The SNMU solution of matrix U can be displayed in a heatmap. The heatmap shows the relative contribution of each substance to a component.

Figure 69 SNMU: matrix \(U\), substance contributions to components.

In Figure 70 and Figure 71 the sparsity parameter is set to 0.1 (not sparse) and 0.8 (sparse), respectively. This leads to components containing different number of substances.

Figure 70 Co-exposure of substances. Heatmap for a solution with 6 components. The sparsity = 0.1. Each component, especially the first, contains many substances (see also Figure 71).

Figure 71 Co-exposure of substances. Heatmap for a solution with 6 components. The sparsity = 0.8. Components contain less substances compared to Figure 70.

In Figure 72, the relative contributions of the substances to the first component are displayed in a piechart.

Figure 72 Relative contributions substances to component 1. The sparsity is set to 0.8 (sparse), estimated sparsity = 0.62.

As mentioned before, one of the nice features of the SNMU algorithm is its recursive character which results in identical components. In Figure 73, the U matrix is visualized using three components. Compare this solution with Figure 71, the first three components are identical. Because of the ordering the plots look slightly different, but a closer inspection of the first 3 components of each solution shows that they are the same. In both figures, component 1 contains mbzp, A, B, C and D; component 2, mibp, E, F and ohmehp; and component 3 mnbp and G.

Figure 73 Heatmap for solution with 3 components. The first 3 components in Figure 73 and Figure 71 contain the same substances.

In paragraph network analysis, an alternative to the SNMU approach is proposed.

For selection of individual(day) exposures with a maximum cumulative ratio above a cutoff and/or above a cutoff percentage in the set of individual(day)s ranked on total exposure, see ‘Cutoff MCR’ and ‘Cutoff percentage’ settings.