[R] Constructing correlation matrices (follow up)

Horace Tso Horace.Tso at pgn.com
Tue Jul 31 01:12:38 CEST 2007

Greg, in light of Doug Bates' question, what i have suggested a little early in response to your question is known as a Wishart matrix with n degree of freedom, which is guarenteed to be positive definite. If this is not what you want, you have to be more specific about the property of this correlation matrix you want to simulate.



I take it that you're trying to generate a random correlation matrix, so first create a covariance matrix,

p = 6
v = matrix(rnorm(p*p), ncol=p)
cov = t(v) %*% v

Then convert it to a correlation matrix,




>>> Gregory Gentlemen <gregory_gentlemen at yahoo.ca> 7/29/2007 7:31:36 PM >>>

I have a seemingly simple task which I have not been able to solve today and I checked all of the help archives on this and have been unable to find anything useful. I want to construct a symmetric matrix of arbtriray size w/o using loops. The following I thought would do it:

p <- 6
Rmat <- diag(p)
dat.cor <- rnorm(p*(p-1)/2)
Rmat[outer(1:p, 1:p, "<")] <- Rmat[outer(1:p, 1:p, ">")] <- dat.cor

However, the problem is that the matrix is filled by column and so the resulting matrix is not symmetric.

I'd be grateful for any adive and/or solutions.


	[[alternative HTML version deleted]]

R-help at stat.math.ethz.ch mailing list
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html 
and provide commented, minimal, self-contained, reproducible code.

More information about the R-help mailing list