| Why does shrinkage help? |
|
|
|
|
|
|
|
|
|
| • | Suppose you have
an unbiased estimator for the |
|||||||
| price of corn and
an unbiased estimator for the |
||||||||
| number of fouls
committed by the leafs. |
||||||||
| • | You can improve
each estimate by taking a |
|||||||
| weighted average
with the other: |
||||||||
| • | For some positive
epsilon, this estimate will have |
|||||||
| a smaller squared
error (but it will be biased). |
||||||||