Hi,

You right, there is no chis_cdf.m and others Statistics function in this Toolbox, the author supposes we already have them.

In case you don't have Matlab toolbox.

Here a Matlab Statistics toolbox that you can integrate :

99380.rar
大小:(2.17 MB)

 马上下载

You'll need a Matlab Optimisation Toolbox too :

99381.rar
大小:(1.79 MB)

 马上下载

[此贴子已经被作者于2007-3-15 22:14:59编辑过]

请问哈，matlab中能不能进行FIGARCH的参数估计啊？

thanks so much

下载了stats和optim，还是找不到 chis_cdf.m文件，能指点一下吗？

在运行demo的第2段时出现了错误，请问各位高手是什么原因呢？

??? Error using ==> max
Matrix dimensions must agree.

Error in ==> finitedifferences at 86
CHG = sqrt(eps)*sign(xCurrent).*max(abs(xCurrent),abs(typicalx));

Error in ==> optim\private\nlconst at 280
[gf,gnc,NEWLAMBDA,OLDLAMBDA,s]=finitedifferences(XOUT,x,funfcn,confcn,lb,ub,f,nc, ...

Error in ==> fmincon at 512
[X,FVAL,lambda,EXITFLAG,OUTPUT,GRAD,HESSIAN]=...

Error in ==> fattailed_garch at 198
[parameters, LLF, EXITFLAG, OUTPUT, LAMBDA, GRAD] = fmincon('fattailed_garchlikelihood', startingvals ,sumA , sumB ,[] , [] , LB , UB,[],options, data, p , q, errortype, stdEstimate^2, T);

Error in ==> ucsd_garch_demo at 98
[parameters, likelihood, stderrors, robustSE, ht, scores] = fattailed_garch(x , 1 ,1 , 'NORMAL');

请帮忙，发现错误

Beginning the UCSD_GARCH demo program. This program is designed to
help you get a basic feel for the programs included, but does not show all
functionality. Please press a key to continue.

We will begin by simulatig a simple univariate GARCH process with normal innovations.
The command to do this is:

[x,h]=garchsimulate([.01; .2; .75],1,1,1000)

where x is the output data series, h is the output conditional variances, the vector [.01; .04; .94]
represents the constant, arch and garch parameters, the two 1s represent the arch and garch orders.
and 500 is the amount of output desired.

We shoudl test out data for the presense of GARCH. This can be done by calling lmtest2
[statistic, pval] = lmtest2(x,5)
where statistic is the estimated statistic and pval are the associated probabilities the squares of the residuals are independent.
??? Undefined function or method 'chis_cdf' for input arguments of type 'double'.

Error in ==> lmtest2 at 40
results.pval=1-chis_cdf(statistic,(1:q)');

我运行demo也出现了同样的错误，我的matlab的版本是R2007a。

我也尝试的修改了一下chi_cdf那个文件lmtest2.m，在matlab里面的stats里面有一个chi2cdf，不过好像还是有问题，难道这个版本的只能在matlab6里面使用？

请高手指点！

我运行demo也出现了同样的错误，我的matlab的版本是R2007a。

我也尝试的修改了一下chi_cdf那个文件lmtest2.m，在matlab里面的stats里面有一个chi2cdf，不过好像还是有问题，难道这个版本的只能在matlab6里面使用？

请高手指点！

太感谢了！

我的安装好了

http://www.kevinsheppard.com/wiki/UCSD_GARCH

写得十分清楚

安装整个JPl toobox

里面包括 ucsd_garch toolbox

然后set path 就可以了

不同意poorsol的说法。

看来poorsol也没有好好看Kevin Sheppard写的文档。

ucsd_garch toolbox的安装绝对没有那么简单。JPL toolbox里面本身的确是有ucsd_garch toolbox，但是JPL toolbox 里面的ucsd_garch toolbox是Kevin Sheppard早期编写的代码，Kevin Sheppard后来逐步发现并且改正了其中的许多错误，所以Kevin Sheppard于2007年4月推出了新版本的ucsd_garch toolbox，版本号是2.0.10 21。Kevin Sheppard早已说过不要使用JPL toolbox里面本身带的ucsd_garch toolbox，因为错误较多。

正确的方法是用Kevin Sheppard于2007年4月推出的新版本的ucsd_garch toolbox，代替JPL toolbox本身自带的ucsd_garch toolbox，替代过程中需要做一些技术处理，保留一些文件而删除一些文件。对于matlab6.5,matlab2007和matlab2008,只要matlab版本不同，对ucsd_garch toolbox所做的修改也不相同。

正确的方法看：正确安装ucsd-garch工具包https://bbs.pinggu.org/thread-376343-1-1.html

[此贴子已经被作者于2008-10-28 22:17:36编辑过]

还有一个人的主页上也有很多关于GARCH模型的toolbox,有一些就是基于USCD的,大家可以看看

http://www.economics.ox.ac.uk/members/andrew.patton/code.html