R devel - Jul 2006 - noncentral F-distributed random numbers (PR#9055)

Full_Name: Long Qu
Version: 2.3.1
OS: Windows XP
Submission from: (NULL) (64.113.93.235)


The QQ-plot of two versions of simulating noncentral F-distributed random
numbers has quite different scales: 
> qqplot(rf(1000,2,15,3),qf(runif(1000),2,15,3))
The rf() function reads:
> rf
function (n, df1, df2, ncp = 0) 
{
    if (ncp == 0) 
        .Internal(rf(n, df1, df2))
    else rchisq(n, df1, ncp = ncp)/rchisq(n, df2)
}
<environment: namespace:stats>

where I believe both the numerator and the denominator should be divided by
their corresponding degrees of freedom. 


My suggested (slighly augmented) version is:
> rffunction (n, df1, df2, ncp1 = 0, ncp2=0) 
{
    if (ncp1 == 0 && ncp2==0) 
        .Internal(rf(n, df1, df2))
    else if (ncp2==0) 
          (rchisq(n, df1, ncp = ncp1)/df1)/(rchisq(n, df2)/df2)
    else  (rchisq(n, df1, ncp = ncp1)/df1)/(rchisq(n, df2, ncp=ncp2)/df2)

}

which incorporated both singly and doubly noncentral F-distribution. 

> version
               _                         
platform       i386-pc-mingw32           
arch           i386                      
os             mingw32                   
system         i386, mingw32             
status                                   
major          2                         
minor          3.1                       
year           2006                      
month          06                        
day            01                        
svn rev        38247                     
language       R                         
version.string Version 2.3.1 (2006-06-01)



Regards, 
Long Qu

On Sat, 1 Jul 2006, longor at iastate.edu wrote:
>
> The rf() function reads:
>> rf
> function (n, df1, df2, ncp = 0)
> {
>    if (ncp == 0)
>        .Internal(rf(n, df1, df2))
>    else rchisq(n, df1, ncp = ncp)/rchisq(n, df2)
> }
> <environment: namespace:stats>
>
> where I believe both the numerator and the denominator should be divided by
> their corresponding degrees of freedom.
>
Yes. Fixed for 2.4.0.

I didn't add a denominator non-centrality parameter, since we don't have
code for qf, pf, or df for the doubly non-central distribution.

 	-thomas