topic Re: Sum in SAS Procedures

Sum

BaileyY — Fri, 19 Dec 2014 18:08:00 GMT

Hi-

I am trying to add individual records and roll up to office, regional and state levels. I was producing multiple table frequencies but the totals are appearing on the right hand side when I am wanting the totals to be within the table. My data looks like this and I've also included my desired results. The regional level rows are the sums of individual office levels. Any help would be appreciated. Thank you.

ID	RACE	GENDER	PMT_TYPE	ICD	VISIT	OFFICE
AA	H	M	HMO	290	IN	9
AB	H	M	HMO	290	OUT	9
AC	H	M	FFS	290	IN	9
AD	H	M	MC	290	IN	9
BA	H	F	HMO	290	OUT	9
BB	H	F	FFS	290	IN	9
BC	H	F	PPO	290	IN	9
BD	H	F	MC	290	IN	9
CA	W	M	HMO	290	OUT	9
CB	W	M	FFS	290	IN	9
CC	W	M	PPO	290	IN	9
CD	W	M	MC	290	OUT	9
CD	W	M	MC	290	OUT	9
DA	W	F	HMO	290	OUT	9
DB	W	F	FFS	290	IN	9
DC	W	F	PPO	290	OUT	9
DD	W	F	MC	290	OUT	9
ZA	H	M	HMO	290	IN	8
ZB	W	F	HMO	290	IN	8
ZC	W	F	HMO	290	OUT	8

Desired output:
LEVEL	RACE	GENDER	HMO	FFS	PPO	MC	ICD290	IN-VISIT	OUT-VISIT
OFFICE9	H	M	2	1	0	1	4	2	1
OFFICE9	H	F	1	1	1	1	4	3	1
OFFICE9	W	M	1	1	1	2	5	2	3
OFFICE9	W	F	1	1	1	1	4	1	3
OFFICE8	W	F	1	0	0	0	1	1	0
OFFICE8	W	F	1	0	0	0	1	0	1
REGION	H	M	2	1	0	1	4	2	1
REGION	H	F	1	1	1	1	4	3	1
REGION	W	M	1	1	1	2	5	2	3
REGION	W	F	3	1	1	1	6	2	4

Re: Sum

data_null__ — Fri, 19 Dec 2014 20:38:26 GMT

This is close.

data office;
   infile cards expandtabs; 
   input (ID  RACE  GENDER   PMT_TYPE ICD   VISIT OFFICE)($);
   cards; 
AA H  M  HMO   290   IN 9
AB H  M  HMO   290   OUT   9
AC H  M  FFS   290   IN 9
AD H  M  MC 290   IN 9
BA H  F  HMO   290   OUT   9
BB H  F  FFS   290   IN 9
BC H  F  PPO   290   IN 9
BD H  F  MC 290   IN 9
CA W  M  HMO   290   OUT   9
CB W  M  FFS   290   IN 9
CC W  M  PPO   290   IN 9
CD W  M  MC 290   OUT   9
CD W  M  MC 290   OUT   9
DA W  F  HMO   290   OUT   9
DB W  F  FFS   290   IN 9
DC W  F  PPO   290   OUT   9
DD W  F  MC 290   OUT   9
ZA H  M  HMO   290   IN 8
ZB W  F  HMO   290   IN 8
ZC W  F  HMO   290   OUT   8
;;;;
   run; 
proc print; 
   run; 
options missing=0; 
proc report nowd list; 
   columns office race gender pmt_type icd visit;
   define office / group descending; 
   define race / group; 
   define gender / group descending; 
   define pmt_type / across order=data;
   define icd / across; 
   define visit / across; 
   run;

Re: Sum

CTorres — Fri, 19 Dec 2014 21:19:06 GMT

try this:

data have;
input ID $ RACE $ GENDER $ PMT_TYPE $ ICD $ VISIT $ OFFICE $;
office=cats('OFFICE',office);
cards;
AA H M HMO 290 IN 9
AB H M HMO 290 OUT 9
AC H M FFS 290 IN 9
AD H M MC 290 IN 9
BA H F HMO 290 OUT 9
BB H F FFS 290 IN 9
BC H F PPO 290 IN 9
BD H F MC 290 IN 9
CA W M HMO 290 OUT 9
CB W M FFS 290 IN 9
CC W M PPO 290 IN 9
CD W M MC 290 OUT 9
CD W M MC 290 OUT 9
DA W F HMO 290 OUT 9
DB W F FFS 290 IN 9
DC W F PPO 290 OUT 9
DD W F MC 290 OUT 9
ZA H M HMO 290 IN 8
ZB W F HMO 290 IN 8
ZC W F HMO 290 OUT 8
;
run;

options missing=0;
proc freq data=have noprint;
tables office*race*gender*pmt_type / out=freq_pmt nopercent;
tables office*race*gender*icd / out=freq_icd nopercent;
tables office*race*gender*visit / out=freq_visit nopercent;
run;
proc transpose data=freq_pmt out=trans_pmt(drop=_name_ _Label_);
by office race gender;
id pmt_type;
var count;
run;
proc transpose data=freq_icd out=trans_icd(drop=_name_ _Label_) prefix=ICD;
by office race gender;
id icd;
var count;
run;
proc transpose data=freq_visit out=trans_visit(drop=_name_ _Label_) suffix=_VISIT;
by office race gender;
id visit;
var count;
run;
data want;
merge trans_pmt trans_icd trans_visit;
by office race gender;
run;

CTorres

Re: Sum

billfish — Fri, 19 Dec 2014 23:02:39 GMT

A solution is the following:

proc sql;

select distinct 'SUM(pmt_type='||''''||strip(pmt_type)||''''||') as '||pmt_type into :p_type separated by ',' from have;

select distinct 'SUM(ICD='||''''||strip(ICD)||''''||') as '||COMPRESS('ICD'||ICD) into :i_type separated by ',' from have;

select distinct 'SUM(VISIT='||''''||strip(VISIT)||''''||') as '||COMPRESS(''''||VISIT||'-VISIT''n') into :v_type separated by ',' from have;

create table want(rename=(off=office)) as
select compress('Office'||put(office,2.)) as off, race, gender,
&p_type , &i_type , &v_type
from have
group by office, race, gender
order by office desc, race, gender desc;

create table want2 as
select 'REGION' as office, race, gender,
&p_type , &i_type , &v_type
from have
group by race, gender
order by race, gender desc;

insert into want select * from want2;
drop table want2;
quit;

Re: Sum

BaileyY — Fri, 19 Dec 2014 23:21:20 GMT

thank you very much. This produced a much closer output I was looking for. Thank you!!!

Re: Sum

Ksharp — Sat, 20 Dec 2014 13:26:41 GMT

SQL can give you less code and faster speed . I am about to make a paper about this when you are building a report .


data office;
   infile cards expandtabs; 
   input (ID  RACE  GENDER   PMT_TYPE ICD   VISIT OFFICE)($);
   cards; 
AA H  M  HMO   290   IN 9
AB H  M  HMO   290   OUT   9
AC H  M  FFS   290   IN 9
AD H  M  MC 290   IN 9
BA H  F  HMO   290   OUT   9
BB H  F  FFS   290   IN 9
BC H  F  PPO   290   IN 9
BD H  F  MC 290   IN 9
CA W  M  HMO   290   OUT   9
CB W  M  FFS   290   IN 9
CC W  M  PPO   290   IN 9
CD W  M  MC 290   OUT   9
CD W  M  MC 290   OUT   9
DA W  F  HMO   290   OUT   9
DB W  F  FFS   290   IN 9
DC W  F  PPO   290   OUT   9
DD W  F  MC 290   OUT   9
ZA H  M  HMO   290   IN 8
ZB W  F  HMO   290   IN 8
ZC W  F  HMO   290   OUT   8
;;;;
   run;

proc sql;
 create table temp as
  select office,race,gender,PMT_TYPE as type ,count(*) as n 
   from office
    group by office,race,gender,PMT_TYPE
 union
  select office,race,gender,ICD as type ,count(*) as n 
   from office
    group by office,race,gender,ICD
 union
  select office,race,gender,VISIT as type ,count(*) as n 
   from office
    group by office,race,gender,VISIT ;
quit;
proc transpose data=temp out=temp1(drop=_:);
 by  office race gender     ;
 id type;
 var n;
run;
proc stdize data=temp1 out=want missing=0 reponly;run;

Xia Keshan