topic Re: Ranuni ( ) in SAS Procedures

Ranuni ( )

KevinC_ — Thu, 15 Oct 2009 13:07:55 GMT

Hello,

I hope I posted this in the right forum. I am trying to figure out how RANUNI works in the following pgm.

Data BB (drop=ran1);
set AA;
ran1=ranuni(1234);
if ran1 < 0.50 then do;
test_cell = 'PM400';
end;

else if ran1 >= 0.50 then do;
test_cell = 'PM401';
end;
run;

Proc Freq data = BB; tables test_cell;

** here is the part of the output that I dont understand*****

test_cell frequency percent
------------------------------------------------
PM400 4194 50.65
PM401 4087 49.35

I understand RANUNI assignes a random number to a variable, in this case 'ran1'. But how does the value of 'ran1' control what records get what test_cell?

Thank you so much for any input !

Re: Ranuni ( )

KevinC_ — Thu, 15 Oct 2009 13:35:09 GMT

let me rephrase my question a little:

according to the result (50.65% vs 49.35%) I can see that the value of 'ran1' was evenly distributed between < 0.50 and >= 0.50. Why is that? Is this by design?

Thanks!

Re: Ranuni ( )

Doc_Duke — Thu, 15 Oct 2009 14:34:34 GMT

Yes, it is by design. RANUNI() is one of many pseudo random number generators in SAS. In particular, RANUNI is designed to generate pseudo random numbers on the real line between 0 and 1 with a uniform distribution.

From the 9.1.3 documentaiton, "The RANUNI function returns a number that is generated from the uniform distribution on the interval (0,1) using a prime modulus multiplicative generator with modulus 231- and multiplier 397204094 (Fishman and Moore 1982) (See References)."

Re: Ranuni ( )

data_null__ — Thu, 15 Oct 2009 14:49:29 GMT

To me the RANTBL function makes this much easier.

[pre]
Data BB;
/*set AA;*/
do id = 1 to 100;
*ran1=ranuni(1234);
*if ran1 < 0.50 then test_cell = 'PM400';
*else if ran1 >= 0.50 then test_cell = 'PM401';

ran1 = rantbl(1234,.5);
length test_cell $5;
test_cell = scan('PM400 PM401',ran1);
output;
end;
run;
Proc Freq data = BB; tables test_cell;
run;
[/pre]

Re: Ranuni ( )

KevinC_ — Thu, 15 Oct 2009 15:20:03 GMT

Thank you both for your input !