Weakly Informative Prior
A prior that deliberately rules out absurd parameter values while staying flat over the plausible range — e.g. Normal(0, 2.5) on a standardized regression coefficient — regularizing estimates without dominating the likelihood. Checked in practice by simulating from the prior predictive distribution.
This Concept is waiting for its first lesson!
A prior that deliberately rules out absurd parameter values while staying flat over the plausible range — e.g. Normal(0, 2.5) on a standardized regression coefficient — regularizing estimates without dominating the likelihood. Checked in practice by simulating from the prior predictive distribution.
Are you a teacher? Sign in to start contributing.
Sign In