Prediction Penalized Analysis of Stroke Based on New Bayesian Hierarchical Priors Model

Abstract

The penalized or the regularization methods nowadays are the most statistical popular tools used for model selection and variable selection procedure. The quality of the regression parameter estimates depends on the prediction accuracy of the estimated model and the model interpretability. Penalized operator methods usually produced the most parsimonious model (less number of predictor and more explanation). This paper utilized new scale mixture of Rayleigh mixing with normal density to study the relationship between the stroke size and some predictors. New hierarchical priors model has developed as well new Gibbs sampling algorithm. The results demonstrated that the proposed model is comparable to some exists regularization models.