12) Choosing priors in Bayesian analysis

Question 1

What are two useful ways to choose a prior in Bayesian analysis

Answer 1

Use a weakly informative prior:
* Have a vague idea of parameter values and use a prior with moderate variance.
* Ensures the prior does not dominate the data, resulting in a weakly regularized estimator

Empirical Bayes methods:
* Determine hyperparameters from observed data
* One approach is to tune the shrinkage parameter λ to minimize MSE