The Smoothly Clipped Absolute Deviation (SCAD) penalty variable selection regularization method for robust regression discontinuity designs

It is necessary to find or search for a way by which the important variables are selected be included in model studied. especially when study data suffers from cut-off point that occurs as result of an abnormal interruption phenomenon studied, leads division experimental units into two groups, where this gap Or jump values observations response variable, so we propose paper new method process estimating and selecting combining Regression Discontinuity Designs (RDD) with (Smoothly Clipped Absolute Deviation (SCAD)) Penalty method. Local linear regression (LLR) was used estimate effect processing on region within optimum bandwidth selection RDD design obtain best model, since (LLR ) basis ( . Three methods were determine IK (Iembens kalyanman) bandwidth, cross-validation (CV) method, The CCT (Calonico, Cattaneo & Titiunik) bandwidth. problem (RDD causal effects treatment estimated using covariates improve efficiency. Where small number observations. Therefore, aims employ (SCAD one variable accuracy covariates. A simulation conducted investigate performance proposed mean squared errors (MSE) choose model. To illustrate use SCAD RDD, R program used..