Predictive and Causal Modeling with the SUPERLEARNER Procedure in SAS® Software

Music City Center 

Room: CC-108 

This workshop introduces the super learner ensemble modeling algorithm as it is implemented in the SUPERLEARNER procedure in SAS software. The super learner uses cross-validation to assess the performance of a set of predictive models and to determine the optimal weights for combining them. It enables investigators to incorporate a diverse set of predictive models into their analysis to achieve better performance. Applications where the super learner has been successful include prediction of patient outcomes and integration with doubly robust methods for causal analysis. The workshop uses worked examples to illustrate how you can use PROC SUPERLEARNER to train super learner models for continuous and binary outcomes, obtain predictions, and compute predictive margins. The workshop also demonstrates how you can use PROC SUPERLEARNER jointly with other SAS procedures for causal effect estimation. Attendees should have a basic familiarity with predictive modeling. After the workshop, they will have a general understanding of the super learner ensemble modeling method and will be able to use PROC SUPERLEARNER to apply the method.