“95% prediction interval”这个概念有个地方不太能理解？

KaronZhang

对于“95% prediction interval”来说，为什么还会有50%的概率少于95%的值落入区间、50%的概率多于95%的值落入区间？

tctcab

信息不足，能再来点背景介绍吗

KaronZhang

tctcab
我看的原版书里是这样表述的：“Here is the precise definition of the 95% prediction interval: Each new observation has a 95% chance of being within the interval. On average, you expect that 95% of the entire population lies within the interval, but there is a 50% chance that fewer than 95% of the values lie within the interval and a 50% chance that more than 95% of the values lie within the interval.”

s609078902

我怎么觉得你表述的有点问题

KaronZhang

s609078902 我看的原版书里是这样表述的：“Here is the precise definition of the 95% prediction interval: Each new observation has a 95% chance of being within the interval. On average, you expect that 95% of the entire population lies within the interval, but there is a 50% chance that fewer than 95% of the values lie within the interval and a 50% chance that more than 95% of the values lie within the interval.”

9-5-2-7

KaronZhang
95%是一个期望概念。
比如，你做100回的重复，每回重复都是进行10000次参数估计。那你不可能在这100回中，每回参数落入置信区间的比例都精确等于9500/10000=0.95。必然是有的比例多于0.95，有的比例小于0.95。这种多于和小于的情况应该是各占50%。

KaronZhang

9-5-2-7 明白了，谢谢您！