Application Of Lasso And Lasso Quantile Regression In The Identification Of Factors Affecting Poverty Levels In Central Java

The condition in which the unknown number parameter to be estimated, p, is much larger than the number of observations, n, is termed high-dimensional. Traditional statistical methods cannot solve high-dimensional problems because they assume many observations and few unknown variables. For high-dimensional modeling, multicollinearity is a frequent phenomenon, causing serious problems with parameter estimation and associated inference and interpretation.. As this reason, Belloni and Chernozhukov in 2011 developed combined methods from Quantile Regression (QR) that is useful for robust regression, and also LASSO that is popular choice for shrinkage estimation and variable selection, becoming LASSO QR. Extensive simulation studies demonstrate satisfactory using LASSO QR in high dimensional datasets that lies outliers better than using LASSO.