[R] generate random numbers from a multivariate distribution with specified correlation matrix

[David Winsemius]

On Aug 23, 2010, at 11:05 PM, rusers.sh wrote:

> Hi,
>  If you see the link http://www.stata.com/help.cgi?drawnorm, and you  
> can
> see an example,
> #draw a sample of 1000 observations from a bivariate standard
> normal distribution, with correlation 0.5.
> #drawnorm x y, n(1000) corr(0.5)
> This is what Stata software did. What i hope to do in R should be  
> similar
> as that.
>  It will be better to only need us to specify the correlation  
> matrix, mean
> values and possible variances. One of my aim is to simulate random  
> fields.
>  Thanks.

?cov2cor

-- 
David.
>
>
> 2010/8/23 Ben Bolker <bbolker at gmail.com>
>
>> rusers.sh <rusers.sh <at> gmail.com> writes:
>>
>>>  rmvnorm()can be used to generate the random numbers from a  
>>> multivariate
>>> normal distribution with specified means and covariance matrix,  
>>> but i
>> want
>>> to specify the correlation matrix instead of covariance matrix for  
>>> the
>>> multivariate
>>> normal distribution.
>>> Does anybody know how to generate the random numbers from a  
>>> multivariate
>>> normal distribution with specified correlation matrix? What about
>>> other non-normal
>>> distribution?
>>
>>  What do you want the variances to be?  If you don't mind that  
>> they're
>> all equal to 1, then using your correlation matrix as the Sigma  
>> argument
>> to the mvrnorm() [sic] function in MASS should work fine.  They  
>> have to
>> be defined as *something* ....
>> If you want multivariate distributions with non-normal marginal
>> distributions, consider the 'copula' package, but be prepared to do
>> some reading -- this is a fairly big/deep topic.
>>
>> good luck.
>>
>

David Winsemius, MD
West Hartford, CT