| To: | <s-news@lists.biostat.wustl.edu> |
|---|---|
| Subject: | Simulation of multivariate student-t distribution |
| From: | "Chen Sichong" <sichong.chen@postgrad.manchester.ac.uk> |
| Date: | Tue, 22 Nov 2005 01:07:58 +0800 |
| Thread-index: | AcXuvh8W3yuKMCT5SMW14pnYNgD13w== |
|
Dear All I have some questions about simulation of multivariate student-t distribution. I am a SPlus 7.0 & FinMetrics 2.0 User
1. Do you know any function in Splus to generate a multivariate student-t distribution? I find rmvnorm for multinormal (Gaussian) but did not find for student t. In MatLab the function is mvtrnd.
2. The univariate series simulated from the multivariate student-t distribution should share the same degree of freedom as multivariate-t. But in my case, I want to release this assumption. What I want is to get two uniform series which should be the probability of univariate of the simulated multivariate series. So is there any way to generate two univariate probability series, whose quantile is multivariate student-t. Or I just can get these probabilities just from the cdf of the two quantile series from the simulation of multivariate student-t.
3. The last question is, if I assume the correlation between the two univariate series changes over time, which is not constant all the time. Is there any way to simulate a multivariate series which captures the characteristic of this changing correlation?
Any suggestion is appreciated.
Many thanks,
SPlus 7.0 & FinMetrics 2.0 User Windows XP
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | surrogate splits in tree, melina cappelli |
|---|---|
| Next by Date: | dev.copy on tabbed graphsheet device, andreas . krause |
| Previous by Thread: | surrogate splits in tree, melina cappelli |
| Next by Thread: | dev.copy on tabbed graphsheet device, andreas . krause |
| Indexes: | [Date] [Thread] [Top] [All Lists] |