topic Re: Splitting one variable into multiple based on criteria from another variable in SAS Procedures

Splitting one variable into multiple based on criteria from another variable

Marjolein — Wed, 02 Nov 2016 14:42:38 GMT

Hi,

I have a question which is intuitively not very difficult, but I cannot seem to find a way to do this in SAS.

I have two variables: variable X (categorical: levels A and B), and a continious variable Z (numeric).

I want to create two new variables out of variable Z, with values that depend on the level of X (A or B).

See the accompanying table. Someone knows the appropriate syntax? Thanks a lot in advance!

X	Z	Z1	Z2
A	4.2	4.2
A	3	3
B	7		7
A	5	5
B	8		8
B	8.5		8.5

Re: Splitting one variable into multiple based on criteria from another variable

Shmuel — Wed, 02 Nov 2016 14:55:30 GMT

I understand that Z1 and Z2 are two new variables to create:

Data want;

set have;

if x='A' then Z1=Z; else

if x='B' then Z2=Z;

run;

Re: Splitting one variable into multiple based on criteria from another variable

ballardw — Wed, 02 Nov 2016 14:56:48 GMT

It looks like you want something like:

data want;

set have;

if X = 'A' Then Z1 = z;

else if X='B' then Z2 = z;

run;

Note: If you have a large number of values of X to consider there are different approaches that may work such as use of the IN comparison if the same thing is done for multiple values or a different structure called SELECT if you have multiple values and each generates a different output.

Re: Splitting one variable into multiple based on criteria from another variable

PeterClemmensen — Wed, 02 Nov 2016 15:04:37 GMT

data have;
   input X $ Z;
   datalines;
A 4. 
A 3  
B 7  
A 5  
B 8 
B 8.5
;

Data want;
  set have;
       if   x='A' then Z1=Z; 
       else x='B' then Z2=Z;
run;

Re: Splitting one variable into multiple based on criteria from another variable

RW9 — Wed, 02 Nov 2016 15:28:05 GMT

To me, the quesiton is why. Currently you have a dataset with two columns, one indicates data position (i.e a where=a would show all z1 for instance), and one is the result. There are no missing elements. What you are asking for is a dataset which is twice as big - 4 variables - with various missing elements, but doesn't actually add anything to the data. So why bother, its just taking more room but without any benefit.

Re: Splitting one variable into multiple based on criteria from another variable

Marjolein — Fri, 04 Nov 2016 10:18:52 GMT

I understand your point. The reason I want to do it, is because they should really represent two distinct variables. Therefore, I really prefer to have them in seperate columns, rather than using the BY statement all the time. Or maybe my mind is rather unlogical...