Steps for selecting the predictor(s) 1.  Run Lasso on 1000 bootstrap samples (Bolasso)    1.1.  Select n-3 subjects without replacement at every run    1.2.  For every subject, select randomly either observation from visit 2 or observation from visit 3    1.3.  Run Lasso on the selected sample and save the vector of coefficients β 2.  Select all the variables with coefficient ≠ 0 in ≥ 60% of the cases Steps for bootstrap evaluation of the model 3.  Run fitlm on 1000 bootstrap samples    3.1.  Select n-3 subjects without replacement at every run    3.2.  For every subject, select randomly either observation from visit 2 or observation from visit 3    3.3.  Run fitlm on the selected sample and save β, Adj. R2 and residuals 4.  Calculate the average model coefficients β and their CI, the CI of the mean and the PI for new observations |  |