Good correlations were obtained for calibration and prediction as indicated by correlation coefficients >0.9965.
The accuracy of prediction, as indicated by correlations, is shown in Table 3.
For all four PM2.5 components and for all numbers of PLS scores, kriging improved prediction accuracy, as indicated by the R and RMSEP statistics for the selected prediction models corresponding to the best performing PLS-only and PLS + universal kriging models (Table 4).
Therefore, the PoPMuSiC-DB mutation set, which is characterized by more homogenous experimental conditions, is presumably more suitable for our prediction scheme, as indicated by the low RMSE value.
Plots of population and individual predictions against observations showed good agreement between population predictions and data, as indicated by the overlay of unity line and locally weighted scatter plot smoothing (LOESS) line, with a slight underestimation at high concentrations.
The resulting reliability index was low in the main mountainous area of Cyprus, where predictions were extrapolations as indicated by the multivariate environmental similarity surface, but medium to high in the main agricultural areas of the country.
The PWM model prediction score of 0.450 was slightly above the prediction threshold (0.412), as indicated by a posterior probability of 0.719 (0.5 at threshold, Table 1).
Model predictions were tested experimentally as indicated in each table.
Our results on the comparison of the prediction methods have shown that LMM is the most appropriate method for AGB prediction using ALS data, as indicated by RMSECV %, but also by considering its strength of accounting for the complex sampling design of the NAFORMA program.
Similarly, EPSA showed that pathway prediction for these samples, as indicated by Spearman rank correlation, also reflected underlying molecular mechanisms.
