PMC:2846953 / 11692-12461 JSONTXT

{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/2846953","sourcedb":"PMC","sourceid":"2846953","source_url":"https://www.ncbi.nlm.nih.gov/pmc/2846953","text":"Using the imputed analysis data sets the MARS procedure was used to fit models to each of the four BOPH using large, medium, and small predictor sets obtained in Random Forests. This procedure produced closed-form linear regression models that display explanatory power comparable to the best Random Forest models. For each of the four BOPH, a preferred MARS model was selected from the MARS runs based on considerations of goodness of fit and parsimony. The selected models were applied and refit in S-PLUS using the validation data set [28]. Graphical model displays are presented to aid in model interpretation. Applying and refitting models using the validation data set protects against model over-fit that can result from aggressive use of data mining procedures.","tracks":[]}