Re: Data Simulation from Correlation Matrix
- From: Gottfried Helms <helms@xxxxxxxxxxxxx>
- Date: Fri, 08 Dec 2006 04:07:27 +0100
Am 07.12.2006 23:46 schrieb Mike - EMAIL IGNORED:
Can someone indicate where I can find the
mathematics and algorithm needed to generate
a simulated data matrix given a correlation
matrix and a random number generator?
Google led me to programs that will do it,
but no adequate explanation.
Thanks for your help.
Mike.
Correlationmatrix (R) ---> (cholesky-decomposition) --> "loadings"-matrix (L)
loadingsmatrix * randomdata ---> correlated data
Say
L*L' = R
and your randomdata are as rowvectors
f1 ........... data ......
F = f2 ............data ......
f3 ............data ......
where F*F' = I ( Identity-mtrix, all correlations=0)
then
z1 ......... data ........
L * F = Z = z2 ......... data ........
z3 ..........data ........
where
Z * Z' = R
The reason is:
Z * Z' = R
(L*F) * (L*F)' = R
L *F*F' *L' = R
L * I *L' = R
L * L' = R
To get L from a given R, you apply cholesky-decomposition
L = cholesky(R)
Gottfried Helms
.
- References:
- Data Simulation from Correlation Matrix
- From: Mike - EMAIL IGNORED
- Data Simulation from Correlation Matrix
- Prev by Date: Re: are there different ARMA processes with the same ACF?
- Next by Date: Re: Bayes inference for comparing stock portfolios
- Previous by thread: Re: Data Simulation from Correlation Matrix
- Next by thread: Re: Data Simulation from Correlation Matrix
- Index(es):
Relevant Pages
|