m******g 发帖数: 51 | 1 最近一个投资公司面试题目(Monte Carlo Simulation):
Suppose we have 10000 gold market daily return data. Let's set a risk
indicator, X (or threshold), to choose 1000 data points which characterized
by lower returns. Then we randomly simulate 10000 data points from the
original data set again and again. For each simulated 10000 data points, we
use X to choose 1000 data points and calculate the average of the 1000
points. Obviously, we can generate a large number of the average numbers by
this way. However, we find the average of original 1000 data chose by X is
deviated to the mean of simulated average numbers as described above.
上面这个方法有什么问题啊?还有请教一下怎么复习Monte Carlo Simulation 有什么
比较好的书? 十分感谢 | k*******d 发帖数: 1340 | 2 没看懂题目
什么叫Then we randomly simulate 10000 data points from the
original data set again and again.
从中选10000个?原来就是10000个啊。。。 | J*****n 发帖数: 4859 | 3
characterized
we
by
这个,你把问题简化一下,换成10个点,找最小的 (原题是10000个,取1000个最小的
平均值)。
那么问题就变成,X1,...,Xn服从1。。10的even dist,那么可以算出来,min(X1) = 1
+ 10/11 = 21 / 11。所以,在给定sample大小的情况下,这个estimator is biased.
【在 m******g 的大作中提到】 : 最近一个投资公司面试题目(Monte Carlo Simulation): : Suppose we have 10000 gold market daily return data. Let's set a risk : indicator, X (or threshold), to choose 1000 data points which characterized : by lower returns. Then we randomly simulate 10000 data points from the : original data set again and again. For each simulated 10000 data points, we : use X to choose 1000 data points and calculate the average of the 1000 : points. Obviously, we can generate a large number of the average numbers by : this way. However, we find the average of original 1000 data chose by X is : deviated to the mean of simulated average numbers as described above. : 上面这个方法有什么问题啊?还有请教一下怎么复习Monte Carlo Simulation 有什么
| C***m 发帖数: 120 | 4 很好的例子。这个bias up或者down和取值1-10有关系吗,还是任意distribution,
bias都会偏向一个方向?谢谢。
1
【在 J*****n 的大作中提到】 : : characterized : we : by : 这个,你把问题简化一下,换成10个点,找最小的 (原题是10000个,取1000个最小的 : 平均值)。 : 那么问题就变成,X1,...,Xn服从1。。10的even dist,那么可以算出来,min(X1) = 1 : + 10/11 = 21 / 11。所以,在给定sample大小的情况下,这个estimator is biased.
| m******g 发帖数: 51 | 5 谢谢!
那怎么样生成这类sample 才能使这个estimator 不是biased的?
1
【在 J*****n 的大作中提到】 : : characterized : we : by : 这个,你把问题简化一下,换成10个点,找最小的 (原题是10000个,取1000个最小的 : 平均值)。 : 那么问题就变成,X1,...,Xn服从1。。10的even dist,那么可以算出来,min(X1) = 1 : + 10/11 = 21 / 11。所以,在给定sample大小的情况下,这个estimator is biased.
| s*******0 发帖数: 3461 | 6 ross 的simulation 浅显易懂
还有就是 经典的 monte carlo method in financial engineer
这个有点深 |
|