TA的每日心情 | 擦汗 2022-2-6 02:55 |
---|
签到天数: 138 天 连续签到: 1 天 [LV.7]常住居民III
|
发表于 2018-11-8 12:39:02
|
显示全部楼层
Shrinkage
- after regression, it is possible to obtain more precise, but slightly biased, estimators of the coefficients, by making them a little bit smaller than the actual value
- eta- and epsilon- shrinkage
Baseball Paradox
- in traditional statistical theory it can be proved that no other estimation rule is uniformly better than the observed average
- Stein shows that sometimes there is a procedure that is better than simply extrapolating from the individual separate averages, no matter what the true batting abilities of the player may be.
- Calculate the average of the average (y bar)
- Shrinking of all the individual average toward this grand average - reduced if greater than, increased if less than (z) - better estimates of the true batting ability than the individual batting averages
- z=y bar + c(y-y bar)
- Constant c is the shrinking factor; c=1, then y=z.
|
|