关于CDO定价 通过matlab实现方法或者其他更好的计算方法

请教各位大神：
如何通过matlab进行CDO定价的实证计算，编不出代码啊。或者是有更好的计算方法。
谢谢了

So would you please provide more details of what you want to do.

such as how many companys in the pool, how many tranches, default intensity, using what copula to model the correlation (Gaussian) and so on...
AND the most important, what problems you are  facing with? You know what you are doing but we don't. You ask for "other" and "better" ways but you didn't give the way you are currently using ^_^. So please provide all the information so that we can help you.

best,

Chemist_MZ 发表于 2013-8-6 23:30
So would you please provide more details of what you want to do.

such as how many companys in the ...

你好，我是想通过KMV模型计算出违约概率，然后引入COPULA函数，对我国之前发行的CDO产品进行定价实证。

xudong1207 发表于 2013-8-20 21:16
你好，我是想通过KMV模型计算出违约概率，然后引入COPULA函数，对我国之前发行的CDO产品进行定价实证。

I am glad you reply 14 days later~

Ok, what's you detailed problem? I think Monte Carlo simulation is very easy to be realized by Matlab. You can first try Gaussian Copula.

best,

Chemist_MZ 发表于 2013-8-20 22:05
I am glad you reply 14 days later~

Ok, what's you detailed problem? I think Monte Carlo simulat ...

不好意思，最近一直很忙，非常谢谢你的耐心解答
我现在算是刚入门，正在尝试计算违约概率
可是中国发行的CDO资产池难以查询到每一笔贷款的具体信息
也就难以获得资产池的违约损失的情况，所以想问问这个地方我应该怎么处理比较合适一点

xudong1207 发表于 2013-8-20 23:26
不好意思，最近一直很忙，非常谢谢你的耐心解答
我现在算是刚入门，正在尝试计算违约概率
可是中国发行 ...

Ok，I see,

I think we can hardly help if you don't have enough data. This is a very big problem for any research. If you really don't have a good data source, you can use the data on the company's balance sheet and its stock price.

After you have each company's default probability from KMV model, the remaining part is to estimate the default correlation. You can check now the default correlation are estimated.

After that, it is easy to do the pricing. Here is a document about KMV by Moody. Hope it will help.
best, KMV-a-Model-D-Risk-2003.pdf.zip (437 KB) 本附件包括：

• KMV-a-Model-D-Risk-2003.pdf

2013-8-21 01:10:41 上传

Chemist_MZ 发表于 2013-8-21 01:02
Ok，I see,

I think we can hardly help if you don't have enough data. This is a very big problem ...

好的，非常谢谢你的解答
我去尝试一下将标的资产的权益市值、长、短期负债等数据整理后，计算出其资产的波动率（德尔塔）和资产价值V。从而计算其违约概率。
这里想请问一些，对于波动率间隔周期取日还是周收益好一些呢？

xudong1207 发表于 2013-8-21 20:33
好的，非常谢谢你的解答
我去尝试一下将标的资产的权益市值、长、短期负债等数据整理后，计算出其资产的 ...

daily

Chemist_MZ 发表于 2013-8-21 21:41
daily

你好，首先谢谢你的耐心解答
这是是我在网上找的一个计算Gaussian copula的代码，但是我运行的时候出现错误。其次我想请问一下，我如何将这个Copula计算出的数据导出，如何是得出相关性关系。谢谢
% this script plot the density of a bivariate gaussian copula function
% pairwise correlation is set at 50%
R = ones(2,2);
r = 0.0242;

R(1,2) = r;
R(2,1) = r;
X = zeros(2,1);
U = zeros(2,1);
gc = zeros(39,39);
h = 0;

for i = 0.025:0.025:0.975
    h = h+1;
    k = 0;
    for j = 0.025:.025:.975
        X =[i,j];
        k = k+1;
        U = norminv(X);
        block1 = 1/(det(R)^0.5);
        block2 = -0.5*U'*(inv(R)-ones(2,2))*U;
        gauss_grid(h,k) = block1*exp(block2);
    end
end
surf(gauss_grid)

运行处来的提示是：??? Error using ==> mtimes
Inner matrix dimensions must agree.

Error in ==> gaussion_copula_density at 21
        block2 = -0.5*U'*(inv(R)-ones(2,2))*U;

讲内层循环里面的X=[i,j] 中间的逗号改成分号就可以了