Fig. 1

a Binomial deviance from the primary analysis of simulated outcome data. Binomial deviance as a function of the regularization parameter (on a logarithmic scale) in the LASSO regression as estimated by cross-validation. The figure contains point estimates as well as error bars signifying the standard error of the estimate. The numbers in the top of the figure denote the degrees of freedom in the model. Vertical dashed lines denote the smallest regularization estimate (leftmost line) and the smallest regularization plus one standard error. b Change in binomial deviance from secondary analysis of simulated outcome data. Out-of-bag (OOB) estimates of the change in binomial deviance as a function of the number of trees in the boosting procedure of the secondary analysis. The relationship is smoothed in the red curve, while vertical and horizontal blue lines demarcate the origin