**EBIR** is an exact Bayesian algorithm applicable to both variable selection and model averaging problems.

This algorithm improves the time complexity of exact inference

with a recursive algorithm
that uses components of one sub-model to rapidly generate

another with a time complexity of *O(m^2)*, where m is the number candidate variables

Computational Statistics and Data Analysis, 56: 1319-1332, doi:10.1016/j.csda.2011.09.026