reBoot

reBoot is a toolbox for statistical calibration of physicochemical property models. In the currently available version, we support polynomial models based on a single scalar input variable. As the model is linear in its parameters, all calibration procedures implemented in reBoot are currently based on different types of linear regression.

For the determination of model prediction uncertainty, we currently provide nonparametric bootstrapping, k-fold cross-validation, and the evidence approximation to Bayesian inference (based on the normal-population assumption).

Note that the statistical methods implemented in reBoot are not limited to single-variable polynomial models that are linear with respect to their parameters. For instance, implementation of nonpolynomial models, many-variable models, or models being nonlinear in their parameters is straightforward.

The reBoot scripts (written in GNU Octave) can be found on our external pageGitHub page. For technical details, please consult the external pagereBoot manual.

We kindly request that, for reproducibility reasons, any use of reBoot in published material should cite the statistical-calibration procedures described in:

J. Proppe and M. Reiher, “Reliable Estimation of Prediction Uncertainty for Physicochemical Property Models”, J. Chem. Theory Comput. 2017

external pageDOI: 10.1021/acs.jctc.7b00235.

The reBoot toolbox is free of charge and distributed under the GNU General Public License v3.