Fig. 2.

Comparison of the true Lasso distribution with the modified residual bootstrap approximation to the distribution of ${\widehat{\beta }}_{\lambda ,1},$ with increasing number of null SNPs. The black curve represents the empirical “true” Lasso distribution of $T_1=\sqrt{n}({\widehat{\beta }}_1-{\beta }_1)$, over 300 simulated datasets. The other curves combine the modified residual bootstrap distribution of $T_1^{\ast \ast }=\sqrt{n}({\widehat{\beta }}_{\lambda ,1}^{\ast \ast }-{\tilde{\beta }}_{\lambda ,1})$ from all 300 datasets