Bayesian fused lasso modeling via horseshoe prior

Abstract Bayesian fused lasso is one of the sparse methods, which shrinks both regression coefficients and their successive differences simultaneously. In this paper, we propose a modeling via horseshoe prior. By assuming prior on difference coefficients, proposed method enables us to prevent over-shrinkage those differences. We also nearly hexagonal operator for with shrinkage equality selection prior, imposes priors all combinations coefficients. Simulation studies an application real data show that gives better performance than existing methods.