[R] [mvtnorm] fast computation of multinormal cdf

[cchevalier]

Hello,
I'm trying to use the pmvnorm function of the mvtnorm to compute the CDF of
the multinormal distribution in rather high dimension (say 100, but 1000
would be nice).

To accelerate the calculations I want to decrease the precision of the
numerical integration,  through the GenzBretz() argument of pmvnorm.

A sample code:

set.seed(8)
Tvect <- rep(1,times=100)
mat <- matrix(c(0.5),nrow=100,ncol=100)
diag(mat) <- rep(1,times=100)
pmvnorm(upper=Tvect,corr=mat,algorithm=GenzBretz(abseps=0.01))

My problem is that the arguments passed to GenzBretz() seem to have no
effect on the precision of the calculation nor on its speed. I tryed to play
with
- maxpts
- abseps
- releps
and the code above gives the same result, as if these arguments were
ignored.

Any ideas? Thanks for your help.




--
View this message in context: http://r.789695.n4.nabble.com/mvtnorm-fast-computation-of-multinormal-cdf-tp4639053.html
Sent from the R help mailing list archive at Nabble.com.