Table 3:
Summary results for the first four pollutant PCs from IMPROVE 2014 data. The R2 have been obtained by leave one out cross validations. The values of variance explained by each PC have been obtained from the whole dataset.
| R2 | Variance explained | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PC1 | PC2 | PC3 | PC4 | PC1 | PC2 | PC3 | PC4 | Total:4PCs | |||
| Traditional PCA GIS PC prediction | 0.55 | 0 | 0.33 | 0.52 | 32% | 17% | 14% | 9% | 73% | ||
| Predictive PCA | 0.55 | 0 | 0 | 0.33 | 32% | 16% | 9% | 12% | 69% | ||
| Predictive PCA GIS PLS prediction | 0.77 | 0.55 | 0.11 | 0.57 | 32% | 16% | 9% | 12% | 69% | ||
| Adaptive predictive PCA | 0.78 | 0.55 | 0.45 | 0.34 | 32% | 17% | 13% | 9% | 71% | ||