The talk will present a general latent variable framework for modelling data features such as heteroscedasticity, skewness, and zero-inflation, as well as a semi-parametric framework for the latent-variable distribution. The talk will discuss the model specifics and estimation, and present simulation results and examples from an American election study and educational measurement.

Collaborators: Camilo Cardenas-Hurtado, Yunxiao Chen, and Giampiero Marra.