topic Re: long to wide for multiple varaibles in SAS Programming

long to wide for multiple varaibles

Ronein — Fri, 03 Jan 2025 19:05:45 GMT

Hello

LEt's say I have a data set called Have with Lonf structure.

I want to convert it to wide structure.

I know to convery long to side for one varaible using proc transpose.

My question-

Here I am doing x6 times proc transpose .

IS there a beter way to do it with less code that tranpsose Long to wide for 6 varaibles together?

Data have'
input CustID YYYYMM monIndex  X W Z R Q Y ;
cards;
111 202401 1 10 11 13 15 19 1
111 202402 2 17 17 12 10 18 0
111 202403 3 19 18 12 12 11 0
111 202404 4 12 12 12 13 14 1
111 202405 5 16 15 14 18 17 1
111 202406 6 16 17 17 18 15 0
222 202401 1 11 11 11 11 11 0
222 202402 2 12 13 14 11 16 0
222 202403 3 19 19 18 16 19 1
222 202404 4 17 17 16 15 18 1
222 202405 5 16 15 16 16 17 0
222 202406 6 14 16 14 13 17 0
;
Run
proc transpose data=have out=wide_1;
by CustID ;
ID monIndex  ;
Var X;
Run;

proc transpose data=have out=wide_2;
by CustID ;
ID monIndex  ;
Var W;
Run;

proc transpose data=have out=wide_3;
by CustID ;
ID monIndex  ;
Var Z;
Run;


proc transpose data=have out=wide_4;
by CustID ;
ID monIndex  ;
Var R;
Run;


proc transpose data=have out=wide_5;
by CustID ;
ID monIndex  ;
Var Q;
Run;

proc transpose data=have out=wide_6;
by CustID ;
ID monIndex  ;
Var Y;
Run;

Data want;
Merge wide_1-wide_6;
by CustID;
Run;

Re: long to wide for multiple varaibles

PaigeMiller — Fri, 03 Jan 2025 19:19:23 GMT

Maxim 19?

Re: long to wide for multiple varaibles

mkeintz — Fri, 03 Jan 2025 19:49:16 GMT

Your WANT dataset is identical to WIDE_6, which presumably is not your intent. Please show what you expect WANT to look like.

Re: long to wide for multiple varaibles

A_Kh — Fri, 03 Jan 2025 20:00:34 GMT

VAR statement accepts multiple variables, no need to transpose each variable separately.

proc transpose data=have out=want prefix=Month_;
by CustID ;
ID monIndex  ;
Var X W Z R Q Y;
Run;

Re: long to wide for multiple varaibles

Tom — Fri, 03 Jan 2025 21:14:29 GMT

Since you have multiple variables your starting point is not truly a TALL dataset.

You should first convert your intermediate dataset into a TALL dataset. Then you can transpose that TALL dataset to a WIDE dataset.

proc transpose data=have out=tall;
  by custid yyyymm monIndex;
  var x -- y ;
run;

proc transpose data=tall  out=want(drop=_name_);
  by custid ;
  id _name_ monindex;
  var col1 ;
run;

Result

If you want the new variables ordered differently add a PROC SORT step in between.

proc sort;
  by custid _name_ monindex;
run;

Result

Re: long to wide for multiple varaibles

Ksharp — Sat, 04 Jan 2025 08:22:52 GMT

Tom already gave you Double Transpose solution.

But if you have a very BIG table,could try MERGE skill proposed by me.

Data have;
input CustID YYYYMM monIndex  X W Z R Q Y ;
cards;
111 202401 1 10 11 13 15 19 1
111 202402 2 17 17 12 10 18 0
111 202403 3 19 18 12 12 11 0
111 202404 4 12 12 12 13 14 1
111 202405 5 16 15 14 18 17 1
111 202406 6 16 17 17 18 15 0
222 202401 1 11 11 11 11 11 0
222 202402 2 12 13 14 11 16 0
222 202403 3 19 19 18 16 19 1
222 202404 4 17 17 16 15 18 1
222 202405 5 16 15 16 16 17 0
222 202406 6 14 16 14 13 17 0
;
proc sql noprint;
select distinct catt('have(where=(monIndex=',monIndex,') rename=(X=X',monIndex,' W=W',monIndex,' Z=Z',monIndex,' R=R',monIndex,' Q=Q',monIndex,' Y=Y',monIndex,'))')
 into :merge separated by ' '
 from have ;
quit;
data want;
merge &merge.;
by CustId;
drop YYYYMM monIndex;
run;

Re: long to wide for multiple varaibles

Ksharp — Sat, 04 Jan 2025 08:38:35 GMT

If you don't have gap in monIndex variable and dataset has been sorted by CustID and monIndex, Could try proc summary solution:

Data have;
input CustID YYYYMM monIndex  X W Z R Q Y ;
cards;
111 202401 1 10 11 13 15 19 1
111 202402 2 17 17 12 10 18 0
111 202403 3 19 18 12 12 11 0
111 202404 4 12 12 12 13 14 1
111 202405 5 16 15 14 18 17 1
111 202406 6 16 17 17 18 15 0
222 202401 1 11 11 11 11 11 0
222 202402 2 12 13 14 11 16 0
222 202403 3 19 19 18 16 19 1
222 202404 4 17 17 16 15 18 1
222 202405 5 16 15 16 16 17 0
222 202406 6 14 16 14 13 17 0
;
proc sql noprint;
select max(n) into :n
 from (select count(*) as n from have group by CustID);
quit;
proc summary data=have nway;
class CustID;
output out=want(drop=_type_ _freq_) idgroup(out[&n] (X W Z R Q Y)=);
run;

Re: long to wide for multiple varaibles

Ronein — Sun, 05 Jan 2025 18:58:01 GMT

Very nice solution.

You wrote that if data set is big then this solution is better than using proc transpose?

Can you give a general rule of thumb what is a big data set? (How many rows, how many columns )

For example:

60 million rows and 10 columns is a big data set?

Re: long to wide for multiple varaibles

Ronein — Sun, 05 Jan 2025 19:01:22 GMT

Thanks for all,

I summarize the provided solutions

Data have;
input CustID YYYYMM monIndex  X W Z R Q Y ;
cards;
111 202401 1 10 11 13 15 19 1
111 202402 2 17 17 12 10 18 0
111 202403 3 19 18 12 12 11 0
111 202404 4 12 12 12 13 14 1
111 202405 5 16 15 14 18 17 1
111 202406 6 16 17 17 18 15 0
222 202401 1 11 11 11 11 11 0
222 202402 2 12 13 14 11 16 0
222 202403 3 19 19 18 16 19 1
222 202404 4 17 17 16 15 18 1
222 202405 5 16 15 16 16 17 0
222 202406 6 14 16 14 13 17 0
;
Run


/**Task: Convert Long to Wide*****/

/***WAY1***/
proc transpose data=have out=wide_1 prefix=X;
by CustID ;
ID monIndex;
Var X;
Run;
proc transpose data=have out=wide_2  prefix=W;
by CustID ;
ID monIndex  ;
Var W;
Run;
proc transpose data=have out=wide_3  prefix=Z;
by CustID ;
ID monIndex  ;
Var Z;
Run;
proc transpose data=have out=wide_4  prefix=R;
by CustID ;
ID monIndex  ;
Var R;
Run;
proc transpose data=have out=wide_5  prefix=Q;
by CustID ;
ID monIndex  ;
Var Q;
Run;
proc transpose data=have out=wide_6  prefix=Y;
by CustID ;
ID monIndex  ;
Var Y;
Run;

Data want;
Merge wide_1-wide_6;
by CustID;
DROP _name_;
Run;

/***WAY2-Quick way--VAR statement accepts multiple variables, no need to transpose each variable separately***/
proc transpose data=have out=want prefix=X;
by CustID ;
ID monIndex;
Var X W Z R Q Y;
/*Var X--Y;*/
DROP _name_;
Run;


/***WAY3-Double Transpose solution***/
proc transpose data=have out=Longer;
by custid yyyymm monIndex;
var x -- y ;
run;
proc sort data=Longer;
by custid _name_ monindex;
run;
/*If you want the new variables ordered differently add a PROC SORT step in between*/
proc transpose data=Longer  out=want(drop=_name_);
by custid ;
id _name_ monindex;
var col1 ;
run;


/***WAY4-IF you have very big data set****/
data want;
Merge 
have(where=(monIndex=1) rename=(X=X1 W=W1 Z=Z1 R=R1 Q=Q1 Y=Y1))
have(where=(monIndex=2) rename=(X=X2 W=W2 Z=Z2 R=R2 Q=Q2 Y=Y2)) 
have(where=(monIndex=3) rename=(X=X3 W=W3 Z=Z3 R=R3 Q=Q3 Y=Y3))
have(where=(monIndex=4) rename=(X=X4 W=W4 Z=Z4 R=R4 Q=Q4 Y=Y4)) 
have(where=(monIndex=5) rename=(X=X5 W=W5 Z=Z5 R=R5 Q=Q5 Y=Y5))
have(where=(monIndex=6) rename=(X=X6 W=W6 Z=Z6 R=R6 Q=Q6 Y=Y6))
;
by CustId;
drop YYYYMM monIndex;
Run;



/***WAY4-IF you have very big data set--Using macro var tooo****/
proc sql noprint;
select distinct catt('have(where=(monIndex=',monIndex,') rename=(X=X',monIndex,' W=W',monIndex,' Z=Z',monIndex,' R=R',monIndex,' Q=Q',monIndex,' Y=Y',monIndex,'))')
into :merge separated by ' '
from have ;
quit;
%put &merge;

data want;
merge &merge.;
by CustId;
drop YYYYMM monIndex;
run;

/**Way5-If you don't have gap in monIndex variable and dataset has been sorted by CustID and monIndex, Could try proc summary solution***/
proc sql noprint;
select max(n) into :n
From (select count(*) as n from have group by CustID);
quit;
%put &n;

proc summary data=have nway;
class CustID;
output out=want(drop=_type_ _freq_) idgroup(out[&n] (X W Z R Q Y)=);
run;

Re: long to wide for multiple varaibles

Ronein — Sun, 05 Jan 2025 19:02:28 GMT

you are right.

IT is because I forgot add prefix.

Data have;
input CustID YYYYMM monIndex  X W Z R Q Y ;
cards;
111 202401 1 10 11 13 15 19 1
111 202402 2 17 17 12 10 18 0
111 202403 3 19 18 12 12 11 0
111 202404 4 12 12 12 13 14 1
111 202405 5 16 15 14 18 17 1
111 202406 6 16 17 17 18 15 0
222 202401 1 11 11 11 11 11 0
222 202402 2 12 13 14 11 16 0
222 202403 3 19 19 18 16 19 1
222 202404 4 17 17 16 15 18 1
222 202405 5 16 15 16 16 17 0
222 202406 6 14 16 14 13 17 0
;
Run


/**Task: Convert Long to Wide*****/

/***WAY1***/
proc transpose data=have out=wide_1 prefix=X;
by CustID ;
ID monIndex;
Var X;
Run;
proc transpose data=have out=wide_2  prefix=W;
by CustID ;
ID monIndex  ;
Var W;
Run;
proc transpose data=have out=wide_3  prefix=Z;
by CustID ;
ID monIndex  ;
Var Z;
Run;
proc transpose data=have out=wide_4  prefix=R;
by CustID ;
ID monIndex  ;
Var R;
Run;
proc transpose data=have out=wide_5  prefix=Q;
by CustID ;
ID monIndex  ;
Var Q;
Run;
proc transpose data=have out=wide_6  prefix=Y;
by CustID ;
ID monIndex  ;
Var Y;
Run;

Data want;
Merge wide_1-wide_6;
by CustID;
DROP _name_;
Run;

Re: long to wide for multiple varaibles

Ksharp — Mon, 06 Jan 2025 01:13:55 GMT

Yes.
proc transpose is too slow to sort a big table.
I think a big table at least have 5 million obs and 100 variables .

Re: long to wide for multiple varaibles

Ksharp — Mon, 06 Jan 2025 01:17:10 GMT

Your WAY2 is not right.
The result is different from me and Tom.

Re: long to wide for multiple varaibles

Ronein — Tue, 07 Jan 2025 06:05:32 GMT

Sorry but this is not the final solution, need one more transpose to get in it full wide structure