how to calculate percentile level

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

计算VaR (Vale at risk) 时, 已经知道density function(不是normal 和 t, 查表得不到), 在95% 置信度下,如何计算percentile level? 需要用什嘛软件? 多谢.

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：percentile calculate percent Level Late Level percentile calculate

You can divide the probability support into G (say G=2000) equi-length intervals. Then compute the area since you have the (Empirical ?) density function(curve?). The value for which the cumulative area is 95% (5%, depending whether you start from the left or the right) is the desired value.  

[此贴子已经被作者于2006-5-1 2:29:10编辑过]

已经知道density function还有啥可算的啊,,
如果不知道,,随便弄个gauss kernel评估一下,,就行了..一般软件都直接给出,,在95% 置信度下,
当然2楼说的也可以..

To zhaosweden and hangover:

Thank you two.

I want to use RiskMetrics to calculate VaR and assme the error is mixed normal distribution. The prob density fuction is g(x)=p*f1(u1, segma1) + (1-p) f2(u2, segma2). I will use MLE to estimate the five parameters and get the density function. But it does not make much sense if I only calcuate condition variance; then use VaR=1.65* CONDITIONAL VARIANCE. I also need to use the definition of VaR to calculate percentile level of mixed normal distribution and get the desire value.

zhaosweden is right, but the process is a little bit complex. Could you tell me how to use some specfic software to solve this problem? Does Eviews work? Could you explain in more details?

Many thanks in advance.

I do not know whether Eviews is OK. If you use that software, I guess you must make your own code(within Eviews). When comes to such question, you have to step into the detail.  

[此贴子已经被作者于2006-5-1 19:44:09编辑过]

Will  any other software work? I do need sove this problem quickly.  Aslo I am not good at programming. A big headache!

My suggestion is simple: do not choose the topic you are not able to do. Otherwise it will take too long time or you end up with some rubbish.

Thanks a lot, zhao. You are definitely right.

reply to: Thanks a lot, zhao. You are definitely right.

***************************

Please note that there are more that people can take than taking my word literally.

I found many Master students choose to use some complicated Econometric Modesl in their master thesis. And such model normally asks for making code (either GAUSS, Metlab, c, ...). But they have only very very veyr poor experience with Eviews or SPSS. The thesis is normally required to be completed within 2-3 month. But to learn GAUSS only may takes them 2 month! My words may not be targetted at you. But I hope you can pick up some useful information here.

老大，I have solved this problem using MATHCAD. Since I am not good at programming, I choose to use some simple software. Just use integral to get the desired value. It is very convenient.