topic Data processing question in SAS Programming

Data processing question

bkoksal — Tue, 31 Jan 2012 14:42:18 GMT

Dear SAS Users,

Here is a puzzle.

I have a data set as follows:

symbol	date	X	Y	Z
ABC	1/2/2012	5	.	.
ABC	1/3/2012	7	.	.
ABC	1/4/2012	2	.	.
ABC	1/5/2012	.	20	.
ABC	1/6/2012	.	.	.
ABC	1/9/2012	.	.	.
ABC	1/10/2012	5	.	.
ABC	1/11/2012	11	.	.
ABC	1/12/2012	5		2
ABC	1/13/2012	8	.	.
DEF	1/2/2012	5	.	.
DEF	1/3/2012	7	15	.
DEF	1/4/2012	8	.	.
DEF	1/5/2012	6	.	.
DEF	1/6/2012	.	.	3
DEF	1/9/2012	9	.	.
DEF	1/10/2012	5	.	.
DEF	1/11/2012	.	.	4
DEF	1/12/2012	.	.	.
DEF	1/13/2012	9	.	.

I want to produce a new variable newX such that

1. if Y^=. and X^=. then NewX=X+Y

2. if Z^=. and X^=. then NewX=X*Z

3. if Y^=. and X=. then the value of Y will be added to the next nonmissing X

4. if Z^=. and X=. then the value of Z will be multiplied by the next nonmissing X

The output will be as follows:

symbol	date	X	Y	Z	NewX
ABC	1/2/2012	5	.	.	5
ABC	1/3/2012	7	.	.	7
ABC	1/4/2012	2	.	.	2
ABC	1/5/2012	.	20	.	.
ABC	1/6/2012	.	.	.	.
ABC	1/9/2012	.	.	.	.
ABC	1/10/2012	5	.	.	25
ABC	1/11/2012	11	.	.	11
ABC	1/12/2012	5		2	10
ABC	1/13/2012	8	.	.	8
DEF	1/2/2012	5	.	.	5
DEF	1/3/2012	7	15	.	22
DEF	1/4/2012	8	.	.	8
DEF	1/5/2012	6	.	.	6
DEF	1/6/2012	.	.	3	.
DEF	1/9/2012	9	.	.	27
DEF	1/10/2012	5	.	.	5
DEF	1/11/2012	.	.	4	.
DEF	1/12/2012	.	.	.	.
DEF	1/13/2012	9	.	.	36

How can this be done? Thanks for any help in advance.

INPUT DATA

symbol date X Y Z

ABC 1/2/2012 5 . .

ABC 1/3/2012 7 . .

ABC 1/4/2012 2 . .

ABC 1/5/2012 . 20 .

ABC 1/6/2012 . . .

ABC 1/9/2012 . . .

ABC 1/10/2012 5 . .

ABC 1/11/2012 11 . .

ABC 1/12/2012 5 2

ABC 1/13/2012 8 . .

DEF 1/2/2012 5 . .

DEF 1/3/2012 7 15 .

DEF 1/4/2012 8 . .

DEF 1/5/2012 6 . .

DEF 1/6/2012 . . 3

DEF 1/9/2012 9 . .

DEF 1/10/2012 5 . .

DEF 1/11/2012 . . 4

DEF 1/12/2012 . . .

DEF 1/13/2012 9 . .

OUTPUT DATA

symbol date X Y Z NewX

ABC 1/2/2012 5 . . 5

ABC 1/3/2012 7 . . 7

ABC 1/4/2012 2 . . 2

ABC 1/5/2012 . 20 . .

ABC 1/6/2012 . . . .

ABC 1/9/2012 . . . .

ABC 1/10/2012 5 . . 25

ABC 1/11/2012 11 . . 11

ABC 1/12/2012 5 2 10

ABC 1/13/2012 8 . . 8

DEF 1/2/2012 5 . . 5

DEF 1/3/2012 7 15 . 22

DEF 1/4/2012 8 . . 8

DEF 1/5/2012 6 . . 6

DEF 1/6/2012 . . 3 .

DEF 1/9/2012 9 . . 27

DEF 1/10/2012 5 . . 5

DEF 1/11/2012 . . 4 .

DEF 1/12/2012 . . . .

DEF 1/13/2012 9 . . 36

Data processing question

Tom — Tue, 31 Jan 2012 15:05:42 GMT

Those conditions overlap. You need to define what order you want the operations to happen.

If x=1 ,y=2 and z=4 then the first two rules imply that NewX is either 3 or 4 depending on which of the two rules to use.

Data processing question

Haikuo — Tue, 31 Jan 2012 15:52:43 GMT

Just like Tom has pointed out, your rules are ambiguous. I will just give an example based on your data as is and there are implied rules such as:

1. y and z are mutually exclusive.

1. z rules triumph over y rules.

data have;

infile cards;

input symbol $ date :mmddyy10. X Y Z;

format date mmddyy10.;

cards;

ABC 1/2/2012 5 . .

ABC 1/3/2012 7 . .

ABC 1/4/2012 2 . .

ABC 1/5/2012 . 20 .

ABC 1/6/2012 . . .

ABC 1/9/2012 . . .

ABC 1/10/2012 5 . .

ABC 1/11/2012 11 . .

ABC 1/12/2012 5 2

ABC 1/13/2012 8 . .

DEF 1/2/2012 5 . .

DEF 1/3/2012 7 15 .

DEF 1/4/2012 8 . .

DEF 1/5/2012 6 . .

DEF 1/6/2012 . . 3

DEF 1/9/2012 9 . .

DEF 1/10/2012 5 . .

DEF 1/11/2012 . . 4

DEF 1/12/2012 . . .

DEF 1/13/2012 9 . .

;

data want (drop=_:);

retain _ry _rz;

set have;

_ry=ifn(missing(y),_ry,y);

_rz=ifn(missing(z),_rz,z);

if not missing(x) then

do;

if missing(_rz) then

do;

newx=sum(x,_ry);

call missing(_ry);

end;

else if not missing(_rz) then

do;

newx=x*_rz;

call missing (_rz);

end;

run;

proc print;run;

Kindly Regards,

Haikuo

Data processing question

bkoksal — Tue, 31 Jan 2012 18:49:59 GMT

Sorry for the confusion,

I thought it was clear from the example but apparently not. Y and Z are not nonmissing at the same time. In other words, if y is nonmissing, then Z is missing and vice versa.

So either we have the case

x=1, y=2 and z=.

x=1, y=. and z=4

bülent

Data processing question

Haikuo — Tue, 31 Jan 2012 19:02:19 GMT

If so, my code seems working for you. Have you tried on your real data?

Haikuo

Data processing question

bkoksal — Wed, 01 Feb 2012 09:51:15 GMT

Hi Haikuo,

My question was not about exactly what I needed for my data but I got the algorithm from your answer. After some slight modification of your code, it worked.

Thanks a lot.

bülent