Incorporating prior information when true priors are unknown: An Information-Theoretic approach for increasing efficiency in estimation

•We develop a new way to incorporate priors when true priors are unobserved.•We improve the efficiency of an Information-Theoretic method of inference.•We use sampling experiments to show increase in efficiency for finite samples.•Our new estimator dominates its other competitors.•A relevant statistic for identifying the best model is developed.

We develop a new way to incorporate prior information within an Information-Theoretic (IT) estimation framework. The estimator considers many potential priors and uses a simple statistic to choose the optimal solution. Our method outperforms its competitors for all finite data.