PMC:2846953 / 25063-25791 JSONTXT

{"target":"http://pubannotation.org/docs/sourcedb/PMC/sourceid/2846953","sourcedb":"PMC","sourceid":"2846953","source_url":"https://www.ncbi.nlm.nih.gov/pmc/2846953","text":"We began by applying a general purpose data mining procedure (Random Forest) to find a list of important variables that could plausibly affect the chosen dependant variables and black box models that relate these variables to the four BOPH. Results from the Random Forest procedure were used as a starting point for the MARS algorithm which produced the multiple regression models. The regression models found by MARS were then applied and refit to a validation data set in SPLUS. It is remarkable that (as shown in Table 5) the R-squared statistics generated by the Random Forests, MARS and regression procedures are in good agreement with each other, suggesting that the proposed models will be reproducible in future studies.","tracks":[]}