Create Dataset with DSL (from Fraktal SAS Programming)

What is this?

There are many occasions when data are needed that are not really present. Think of a particular test scenario for generic report design or a series of artificial datasets used to test a specific algorithm.

Representing a fully featured 3rd generation programming language, SAS DSL provides all means to generate data of unlimited variation in structure, size and precision.

This example will make use of three iteration concepts and two random number generators to produce a sample dataset with character, integer and floating-point columns.

Documented Code

Code executed	Function performed
data random;	Initiate data step processing with a DATA statement and a name
do group = 'A','B','C','D','X','Y','Z';	Start outer loop over discrete values given in comma separated list
do seed = 1 to 1000 by 1;	Start inner loop over integers given by start, end and increment
select (group);	Start branching over values from outer loop
when ('A','B','C','D') noise = ranuni(seed);	Assign uniform distribution random number generator to listed values
otherwise noise = normal(seed);	Assign normal distribution random number generator to remainder of values
end;	End branching over values from outer loop
output;	Write current content of PDV to dataset 'random'
end;	Terminate inner loop
end;	Terminate outer loop
run;	End the data step run-group

Results

When issuing the phrase "vt random" in the command window of your SAS session, the generated table will pop-up in a "Viewtable Window". It's easy to recognize that random number properties have changed according to the algorithm coded:

group	seed	noise
D	990	0.25404
D	991	0.28915
D	992	0.58119
D	993	0.22943
D	994	0.28024
D	995	0.94102
D	996	0.27041
D	997	0.86849
D	998	0.40269
D	999	0.47663
D	1000	0.63694
X	1	0.82346
X	2	-1.04528
X	3	2.59510
X	4	-1.61314
X	5	0.78804
X	6	1.24662
X	7	1.31970
X	8	-0.08472
X	9	-1.37055
X	10	-0.20202