Abstract
Selecting active variables for a QR model is difficult. Selecting the right group of predictors often improves prediction accuracy. To improve scientific understanding, choose a smaller subset. Several methods have been presented to find the active subset. Estimating model parameters aims to find the best estimators for accurate predictions. Estimating all the parameters in the high-dimensional data request yields a weak prediction with large correlations between independent variables, resulting in incorrect findings. Variable selection (V.S) is a key challenge in modelling high-dimensional data. Linear QR selection variables and estimation are studied using the Bayesian hierarchical approach. Regularization bridge and ordinal composite quantile regression are our specialities. This work proposes a Bayesian reciprocal adaptive bridge composite quantile regression for ordinal variable selection and estimation. A new Gibbs sampling approach is developed for comprehensive conditional posterior distributions. We look at how Bayesian reciprocal adaptive bridge composite quantile regression for ordinal data (BrABCQRO) stacks up against other Bayesian and non-Bayesian approaches. The posterior, prior, and conditional distributions are all talked about together. For full conditional posterior distributions, a new Gibbs sampling method is created. A real-world example and many simulation examples show that the suggested methods often work better than standard ones.
