Additive Density-on-Scalar Regression in Bayes Hilbert Spaces With an Application to Gender Economics

This talk introduces a novel approach to modeling densities influenced by scalar covariates using structured additive regression models in Bayes Hilbert spaces. This framework handles continuous, discrete, and mixed densities, ensuring nonnegativity and integration to one.

An application to gender economics data explores the distribution of a woman’s income share in a couple, accounting for year, residence, and child age. The model addresses mixed densities, such as single-earner couples, and interprets effects using odds-ratios.

Theoretical results include consistency and asymptotic normality of the Bayes space (penalized) likelihood estimator, with practical estimation linked to Poisson additive models. This methodology advances understanding of complex sociological and economic phenomena.

Organized by:

Department of Statistics LMU Munich