Fig. 2
From: Lasso regularization for left-censored Gaussian outcome and high-dimensional predictors
Mean Square Error (MSE) calculated on test data of size ntest=100 and 200 replications using gold standard method (GoldS), Gaussian Buckley-James (Gaussian BJ) and 1-Step version (Gaussian BJ 1-Step), non-Parametric Buckley-James (NonParBJ) and simple imputation by LOD (LOD). Data were generated to present different censoring rates: 0% (uncensored), 20% (moderate), 50% (high), and 70% (severe), which are represented by different gray degrees