开发者:上海品职教育科技有限公司 隐私政策详情

应用版本:4.2.11(IOS)|3.2.5(安卓)APP下载

AmberDe · 2020年02月22日

问一道题:NO.PZ2020011101000030 [ FRM I ]

问题如下:

If returns are fat-tailed, and we instead assume that they are normally distributed, is the probability of a large loss over-or under-estimated?

解释:

We will underestimate the probability of a large loss because the normal has thin tails. For example, the probability of a 4σ4\sigma loss from a normal is .003% (3 in 100,000). The probability of a 4σ4\sigma loss from a standardized Student’s t4t_4 is .24% or 80 times more likely.

老师请问这道题怎么理解?看不太懂。谢谢
1 个答案

品职答疑小助手雍 · 2020年02月22日

同学你好,肥尾的分布,因为尾巴比较肥,所以比如同样倍数标准差的损失(解析用的4倍举例)对应的累积概率分布比正态分布要高。

其实就相当于两个分布,你在左边尾巴上来一刀,厚尾的分布左边的部分累积概率大。