[R] [FORGED] Problem running mvrnorm

[Rolf Turner]

On 19/11/15 10:54, Gang Chen wrote:
> Thanks a lot for the pointer, Rolf!
>
> You're correct that
>
>   E <- eigen(Sigma,symmetric=TRUE)
>
> does lead to the same error on the RedHat machine. However, the same
> computation on my Mac works fine:
>
> E <- eigen(Sigma,symmetric=TRUE)
>
> E$values

<SNIP>

I used your constrSigma() function to create Sigma and did eigen() to 
it.  In a trice I got results which agree with those that you show.

So it's not *Linux* as such that is the problem.

(a) What happens if you do svd(Sigma)?  Or, in accordance with Peter 
Dalgaard's suggestion, chol(Sigma)?

(b) It would seem that you need to follow the lines of enquiry suggested 
by Peter.

cheers,

Rolf Turner

-- 
Technical Editor ANZJS
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276