Abstarct: We develop Bayesian models for density regression with emphasis on discrete responses. The problem of density regression is approached by considering methods for multivariate density estimation of mixed scale wariables, and obtaining conditional density estimation from the latter. The approach to multivariate mixed scale outcome density estimation that we describe represents discrete variables, either responses or covariates, as continuse latent variables that are thresholded into ‎discrete‎ ones. We present and compare several models for obtaining these thresholds in the challenging context of count data analysis where the response may be over-dispersed or under-dispersed in som of the regions of the covariate space. We utilize a nonparametric mixture of mutivariate Gaussians to model the directly observed and the latent continuouse variables. We present a Markov chain Monte Carlo algorithm for postorior sampling and provide illustrations on density, mean and quantile regression utilizing simulated and real datasets.