Figure 8From: Hierarchical Cluster-based Partial Least Squares Regression (HC-PLSR) is an efficient tool for metamodelling of nonlinear dynamic modelsTest set prediction results from HC-PLSR, PLSR and OLS for the mammalian circadian clock. Histograms of A) R2 values and B) RMSEP values from test set predictions of the mammalian circadian clock state variable time series as functions of the parameters, with HC-PLSR, global polynomial PLSR and global polynomial OLS. In HC-PLSR, 4-19 PLS components were used in the regression models, while in PLSR 3-17 PLS components were used. The mean R2test set over all state variables was 0.99 for HC-PLSR, 0.96 for PLSR and 0.97 for OLS.Back to article page