Figure 5From: Variable selection methods for developing a biomarker panel for prediction of dengue hemorrhagic feverRegularized Regression Path. Shown is the solution path β(s) as a function of s. As we can see, any segment between two adjacent vertical lines is linear, hence the whole solution path is piecewise linear. Another important feature of this solution path is, over a certain range of values of s, only the relevant predictor variables, that is, IL10, lymphocytes, and platelets have nonzero fitted coefficientsBack to article page