Answers are grounded in this paper and cite their sources.
This paper describes variable selection of grid based 3D QSAR investigations. Three steroid data sets were studied using GOLPE like variable selection as well as variable selection with contiguous domains (boxes). The predictivity of the selected variable models of the training sets was in all cases improved compared with the corresponding original models using all variables. However, the predictivity of the test sets was in a few cases only marginally improved by variable selection. Domain-based variable selection did not improve the predictivity of the external test sets.
No takes yet. Share an insight, caveat, or question.
Ulf Norinder (1996) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: