Super Learner and R Coding

Laura Balzer

Super-Size your learning with the following SERplaylist! Begin with some real-life applications of the Super Learner to predict ICU mortality (Pirracchio et al. 2015), optimize strategies for HIV RNA viral load monitoring (Petersen et al. 2016), and forecast violence among inmates (Bacak and Kennedy 2018). Then dive into Luque-Fernandez et al. (2018) to explore the consequences of data-adaptive estimation in Causal Inference. With your curiosity piqued, check out Naimi and Balzer (2018) for a look under the algorithmic-hood.

As a bonus, get down and dirty with computer coding! Try R Lab3, by Petersen and Balzer, to hand-code discrete Super Learner (a.k.a. the cross-validation selector). Then check out Kennedy’s GitHub to “Git” up and running with the SuperLearner R package, including the selection of hyperparameters and multicore parallelization. Finally, hand-code the full ensemble Super Learner to minimize prediction error or maximize the area under the received operator characteristic curve (AUROC) as implemented in Naimi and Balzer (2018).

And with this playlist complete, we can all be Super Learners!
“Super Learner…
It’s our hero…
Going to take bias (due to model misspecification) down to zero!”
{To be sung to the tune of Captain Planet and the Planeteers (1990-1996)}

Leave a Reply