MENTAL MODEL · M3645

Stein's Paradox

DecideHigh supportDecision Science

Included

Updated 2026-07-31

Loading revision record…

INTRODUCTION

English translation pending.

CORE DEFINITION

A result in mathematical statistics published by Charles Stein, with the improved estimator developed with Willard James. Its core proposition is that when estimating three or more independent normal means, the James-Stein estimator, which pulls each estimate toward a common point, achieves lower total squared error than estimating each mean separately, even though the individual estimates are then biased. The key constraint is that the gain is measured in aggregate error across all parameters; any single estimate may be worse, and the result requires three or more dimensions.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

- Pooling decision: when estimating several related quantities, consider shrinking them toward a common value instead of treating each in isolation. - Error accounting: judge the set of estimates by total error rather than by whether each one is individually unbiased. - Weak-signal detection: use cross-parameter information when individual samples are too small to support separate estimates.

anchor

Anchor fast decisions

Each separate estimate carries sampling variance, and allowing bias in exchange for reduced variance can lower total error when the number of parameters is large. Shrinking toward a common point injects a controlled amount of bias but removes more variance than it adds, so aggregate error falls. The reason this seems to violate intuition is that unbiasedness is a property of each estimator considered alone, while the comparison is made over the joint performance of the whole set.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

Identify the set of quantities you are estimating and confirm there are three or more. 2. Estimate each separately and record the variance of each estimate. 3. Compute a pooled or shrunk estimate that pulls each value toward the group center. 4. Compare total error across the whole set, not the accuracy of any single estimate. 5. Recheck whether the pooling assumption still holds if the group is not homogeneous.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Stein's_exampleZH · Explicit
verified

RELATED MODELS