Asymmetric Prior in Wavelet Shrinkage

In bayesian wavelet shrinkage, the already proposed priors to coefficients are assumed be symmetric around zero. Although this assumption is reasonable in many applications, it not general. The present paper proposes use of an asymmetric shrinkage rule based on discrete mixture a point mass function at zero and beta distribution as prior non-parametric regression model. Statistical properties such bias, variance, classical risks associated provided performances obtained simulation studies involving artificial distributed Donoho-Johnstone test functions. Application seismic real dataset also analyzed.