Bayesian estimation with INLA for binary logit mixed models

What is it about?

We perform an extensive simulation study to compare the results of fitting a random intercept logit model with three competing methods: Bayesian INLA, Bayesian Gibbs sampling, and maximum ikelihood with adaptive quadrature.

Featured Image

Why is it important?

The simulations show that INLA has an excellent performance as it achieves good accuracy (similar to MCMC) with reduced computational times (similar to adaptive quadrature). A further finding is that, for Bayesian methods, the specification of the prior distribution for the cluster variance plays a crucial role and it turns out to be more relevant than the choice of the estimation method.