The results of our study demonstrate that model-selection error may introduce 20 to 40% uncertainty into a live-tree carbon estimate, possibly making this form of error the largest source of uncertainty in estimation of live-tree carbon stores.
This approach is differentiated from the usual sequential model development for given impedance spectra by its emphasis on obtaining supporting observations to guide model selection, use of error analysis to guide regression strategies and experimental design, and use of models to guide selection of new experiments.
Global error analysis of different models showed that, independent of model selection, the minimum fit error is achieved when the maximum population of the visible conformer is around 70%%.
Since we also use cross-validation for estimating the performance of the overall method (including the algorithm for selecting λ), this results in two nested cross-validation loops, one for model selection and one for error estimation.
Two are the standard model selection criteria of model error and simplicity and three are based on the expected behavior of hidden system variables.
A bias towards parameter-rich models can occur during the model selection process if measurement error is not considered [ 19].
One of the approaches to distinguish between model structures is to perform model selection using both model error and complexity, i.e., using the E C model selection criterion.
Even genome-scale analyses may be susceptible to systematic error when model selection is omitted or a poor model is chosen, particularly when divergence among genes is high.
Specifically, the model selection step with minimum prediction error variance was chosen in five-fold split-sample cross-validation within the training set, using the 'choose = CV cvMethod = split(5)' option in SAS Proc GLMSelect (SAS Institute 2013).
Actually, AEs were a rather weak predictor that would have been eliminated had we used aggressive search algorithms with cross-validation or other model selection tools that target prediction error (as suggested in [ 21, 22]).
