Figure 2

Individually matched case-control study. The Lasso regularization path as a function of λ for the paired case-control design. The black vertical line indicates the λ value optimizing the cross-validation criterion. Coefficient values of drugs selected are indicated in black, the others are in gray. Only drugs estimated to be risk factors (positive ${\widehat{\beta }}_j$) are displayed. Potential confounders (sex, age, chronic disease), forced in the model, are omitted.