bayesian_shrinkage

Computes Bayesian shrinkage estimate from observed data and prior.

Computes Bayesian shrinkage estimate from observed data and prior. Normal-normal conjugate model: posterior_mean = (n × observed + prior_weight × prior_mean) / (n + prior_weight)

Signature

{{ bayesian_shrinkage(n_col, observed_col, prior_mean_expr, prior_weight) }}

Arguments

Argument	Description
n_col	column name or expression for sample size (e.g., ‘panel_observations_n’)
observed_col	column name or expression for observed mean (e.g., ‘observed_coefficient’)
prior_mean_expr	expression for prior mean (e.g., ‘0’ for a zero-centered prior, or a column ref)
prior_weight	numeric or expression for prior strength (e.g., ‘10’ means 10 equivalent samples)

Returns

scalar expression composable in SELECT lists. Tolerates n=0 by returning NULL (via NULLIF), which is the correct behaviour for unobserved cells.

Usage

SELECT
      driver_id,
      {{ bayesian_shrinkage('sample_count', 'observed_skill', '0', '5') }} as shrunken_skill
    FROM panel_data