Generating unique random values

[Medvedeff]

Much like my last post for help, I have another problem that I just cannot solve despite my best efforts. I need to generate random integers bounded on a specifyed interval, but I need them to all be unique. In order to do this I am generating random numbers from a uniform distribution, scaling them and then rounding them to the desired interval.

My goal, at least with the code that I have shown below, is to generate 7 datasets, each containing 5 unique integer values. I dont care if different datasets contain the same values, but each individual dataset needs 5 unique values.

The code below is only part of what I have written, but it is the most relevant to what I am asking about. I can post more of the code if need be. Also, the code that I have below is set in a macro statement, which later gets called in a different macro.


%macro samp;
DROP N;
DO SAMPLE=1 to &n;
DO N=1 to 5;
X=Uniform(33);
Y=(X*9)+1;
RY=Round(Y);
RY2=RY;
IF SAMPLE=&n THEN OUTPUT &name&n;
END;
END;
%mend samp;


I have tried several different things and just cant quite get a solution. This site was so helpful with my last post that I figured I would try again. Thank you for any help in advance.

 

[Medvedeff]

Thus far I have been trying to do this by creating a lag of the RY variable, sorting by the lagRY and then putting a statement such as:

if lagry=ry then return;

Ive tried various versions of the statement in various places, and nothing seems to work. One of the problems is that I need to nest the sorting procedure inside of the macro, but it wont resolve if I do that.

Ive also been trying to put a macro label in, so that I can have a statement such as:

if lagRY=RY then out;

out: n=n-1

The goal being that if lagry = ry, then to stop in the middle of that iteration and strike it from the record by redoing that n'th iteration.

Any ideas?

 

[dje]

This is the way that I would try to do it. I am not sure I fully understood your desires. If this is not what you need you might be able to add some proc sort nodupkey; commands to filter unwanted duplicates if that is a problem.



Good Luck,
dje



Code below:



data gg;

%macro samp;

DO SAMPLE=1 to 9;
DO N=1 to 5;
X=Uniform(33);
Y=(X*9)+1;
RY=Round(Y);
RY2=RY;output;
END;
END;
%mend samp;

%samp;


data a1;set gg;
where sample=1;


data a2;set gg;
where sample=2;

data a3;set gg;
where sample=3;

data a4;set gg;
where sample=4;

data a5;set gg;
where sample=5;

data kjl;h=4;output;run;