On Latent Variable Models For Bayesian Inference With Stable Distributions And Processes