[R] R code for var-cov matrix given variances and correlations

[Peter Dalgaard]

"Haynes, Maurice (NIH/NICHD)" <haynesm at cfr.nichd.nih.gov> writes:

> If the vector of variances were
> var.vec <- c(14, 12, 7) 
> and the vector of correlations were 
> cor.vec <- c(.4, .2, .5), 
> then the vector of covariances would be:
> > covAB(c(.4, .2, .5),c(14, 14, 12), c(12, 7, 7))
> [1] 5.184593 1.979899 4.582576
> >
> and the variance-covariance matrix with covariances rounded to 
> the first decimal place would be:
> > vmat <- matrix(c(14, 5.2, 2.0, 5.2, 12, 4.6, 2.0, 4.6, 7),
> + nrow=3)
> > vmat
>      [,1] [,2] [,3]
> [1,] 14.0  5.2  2.0
> [2,]  5.2 12.0  4.6
> [3,]  2.0  4.6  7.0
> > 


# First fill in the correlation matrix:

V <- matrix(NA,3,3)
diag(V) <- 1
V[lower.tri(V)] <-  c(.4, .2, .5)
V[upper.tri(V)] <- t(V)[upper.tri(V)]

# then scale rows and columns

D <- diag(sqrt( c(14, 12, 7)))
D %*% V %*% D

# or, more efficiently

s <- sqrt( c(14, 12, 7))
sweep(sweep(V,1,s,"*"),2,s,"*")




-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907