Abstract
AbstractThis paper focuses on Bayesian reciprocal lasso regression with right censored response variable. Choosing the important variables that relevant on the response variable is very common goal of the regression analysis. The reciprocal lasso adds the reciprocal of L1-rorm in the penalty function. Reciprocal lasso is a regularization method that provides variable selection procedure with more interpretation regression model. We employed the scale mixture of double pareto (SMDP) and the scale mixture of truncated normal (SMTN) that proposed by Mallick et al. (2020) and we with somemodification for (SMTN) in the right censored limited dependent variable. New hierarchical prior model and new Gibbs sampler algorithm have developed. Some simulation examples have conducted to analysis the behavior of the posterior distributions. The results show that the employed scale mixture types outperform other common regularization methods in both of the simulation.
