We propose a fully Bayesian inference for semiparametric joint mean and variance models on the basis of B-spline approximations of nonparametric components. An efficient MCMC method which combines Gibbs sampler and Metropolis–Hastings algorithm is suggested for the inference, and the...