Fig. 3

LASSO (Least Absolute Shrinkage and Selection Operator) regression analysis for biomarkers and cognitive decline. The lambdas (x axis) and the corresponding coefficients (y axis) for the different variables are plotted. The curves represent the change of the coefficient for each lambda and decrease with the increase of the penalty term (lambda). The first line is the minimal value of lambda that results in the best cross validation result. The second line is the largest lambda within 1 standard error of the minimal lambda. This lambda corresponds to a higher penalized model.