On May 10, 2010, at 2:55 PM, Sergio Andrés Estay Cabrera wrote:
Dear R users and specially Albyn and Giovanni,
thanks for your answers, but in fact I supposed the same at the
beginning of my problem. However, when I generate the data seldom I
obtain the expected correlation. For example using this code:
fz<-function(n,t,rho){
f<-NULL
for(i in 1:n){
s<-rmvnorm(n=t,mean=c(0,0),sigma=matrix(c(1,rho,rho,
1),ncol=2),method='svd')
paso<-cor(cumsum(s[,1]),cumsum(s[,2]))
f<-c(f,paso)}
f<-f
}
I believe you may want to search on the term "quantos model". If you
have two independent Brownian processes W_1(t) and W_2(t) then W_3 =
rho*W_1(t) +sqrt(1-rho^2)*W_2(t) will theoretically have correlation
rho with W_1. Sampling will of course giving varying instantiation
values.
--
David
citation: "Financial Calculus", Baxter M; Rennie A (1996)
and then plot the histogram of the results, it is possible to
observe that the distribution of the values is asymmetric with most
of the simulations close to 1 when the value of rho is higher than
0.3 and looks like a uniform distribution with values below 0.3.
I suspect than the only possibility is using a brute force algorithm.
Any advice would be helpful
Sergio A. Estay
*CASEB *
Departamento de Ecología
Universidad Catolica de Chile
Albyn Jones wrote:
Sums of correlated increments have the same correlation as the
original
variables...
library(mvtnorm)
X<- matrix(0,nrow=1000,ncol=2)
for(i in 1:1000){
Y <- rmvnorm(1000,mean=mu,sigma=S)
X[i,] <- apply(Y,2,sum)
}
cor(Y)
[,1] [,2]
[1,] 1.0000000 0.4909281
[2,] 0.4909281 1.0000000
So, unless you meant that you want the _sample_ correlation to be
pre-specified, you are all set.
albyn
On Sun, May 09, 2010 at 09:20:25PM -0400, Sergio Andrés Estay
Cabrera wrote:
Hi everybody,
I am trying to generate two random walks with an specific
correlation, for example, two random walks of 200 time steps with
a correlation 0.7.
I built the random walks with:
x<-cumsum(rnorm(200, mean=0,sd=1))
y<-cumsum(rnorm(200, mean=0,sd=1))
but I don't know how to fix the correlation between them.
With white noise is easy to fix the correlation using the function
rmvnorm in the package mvtnorm
I surfed in the web in the searchable mail archives in the R web
site but no references appears.
If you have some advices to solve this problems I would be very
thankful.
Thanks in advance.
Sergio A. Estay
*CASEB *
Departamento de Ecología
Universidad Catolica de Chile
--
“La disciplina no tiene ningún mérito en circunstancias ideales. ”
– Habor Mallow
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
--
“La disciplina no tiene ningún mérito en circunstancias ideales. ” –
Habor Mallow
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
David Winsemius, MD
West Hartford, CT
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
