Consumer surface · principia-connector · HO-085 (PRN-144)

Local Validity Report

Per predictor in your selection battery, fuses the borrowed literature prior with your own predictor→criterion data to produce a posterior local validity. Predictors whose posterior r is indistinguishable from zero are flagged Prune. Predictors where your local data has drifted from the prior or where n is too small to trust are flagged with Decay warnings. The observation weight bar shows how much your own data vs the borrowed prior is driving each estimate.

Pairs with Battery Builder (assemble → validate locally → prune). Enter local r and n per predictor, or supply raw predictor×criterion pairs via the API directly.