Stochastic Algorithms for Bayesian Estimation (of Latent Variable Models), Maarten Marsman

I worked on Rejection and Metropolis type algorithms for Bayesian estimation using the Gibbs sampler. These algorithms are specifically designed for large-scale inference. This means that their efficiency improves as more data are coming in. In the forthcoming article I focused on applications of (complex) Item Response Theory (IRT) models.

Currently, I am working on a Rejection based algorithm using Quadratic Minorization, specifically designed for Bayesian estimation of Ising-type Network models, in combination with the previously designed algorithms. This is part of a line of research that relates Ising-type Network models to IRT models, with important applications in educational measurement.