topic Re: Binomial and Poisson distribution in SAS/IML Software and Matrix Computations

Binomial and Poisson distribution

maureen — Sat, 08 Nov 2014 18:48:47 GMT

Hello

I want to generate a random sample from Binomial distribution (it will be included inside a loop) and another one from Poisson distribution.

How can I do this?

Thanks for your help

Re: Binomial and Poisson distribution

IanWakeling — Sun, 09 Nov 2014 10:09:19 GMT

In SAS/IML there are essentially two ways. The random sample can be generated one at a time with the usual Base SAS functions.

call streaminit(1234);

do i = 1 to 10;

b = rand('binomial', 0.05, 20); /* p=0.05 n=20 */

p = rand('poisson', 5); /* lambda=5 */

print b p;

end;

However the RANDGEN call is usually the best option. In general it is more efficient not to use a loop and generate the random numbers all at the same time.

b = j(10, 1);

p = j(10, 1);

call randseed(4321);

call randgen(b, 'binomial', 0.05, 20);

call randgen(p, 'poisson', 5);

print b p;

Re: Binomial and Poisson distribution

Rick_SAS — Sun, 09 Nov 2014 11:53:39 GMT

Ian's answer is correct. I will add that if you are only going to simulate ONE sample (that is, not inside a loop), you can use the newer RANDFUN function in SAS/IML 12.3: Convenient functions vs. efficient subroutines: Your choice - The DO Loop

Since you mention the binomial and Poisson distribution together, I wonder whether you are forming a compound distribution? If so, see this article: A different way to interpret the negative binomial distribution - The DO Loop