Distribution of observed absolute errors and uncertainties predicted by DPRESS for three different CoMFA models. Projection parameters and color coding are the same as in Fig. 3A except that the horizontal dimension has been compressed somewhat. Symbol size is proportional to the magnitude of the observed error or predicted uncertainty. Compounds from the respective training sets are represented by stars. (A) Observed absolute errors for boosted training set A, which had the best external predictive performance (sPRED = 0.633; sCV = 0.762). (B) Observed absolute errors for boosted training set B, which had the best internal predictive performance (sCV = 0.681; sPRED = 0.637). (C) Observed absolute errors for the biased training set (sCV = 0.489; sPRED = 0.744). (D) Predicted uncertainties for boosted training set A. (E) Predicted uncertainties for boosted training set B. (F) Predicted uncertainties for the biased training set R.