topic Re: query in SAS Procedures

query

rakeshvvv — Mon, 29 Dec 2014 22:02:57 GMT

I have following dataset with the following 4 variables...subjid,number,cat,and visitnum..... the number of lesion must remain constant from visit to visit.....i should output where The number of target lesions is not constant from visit to visit....the below sample subject have constant number from visit to visit and we should output discrepancies if there any like those highlighted in blue.....

100010001 T02 TARGET LESIONS 0

100010001 T01 TARGET LESIONS 0

100010001 T03 TARGET LESIONS 0

100010001 T02 TARGET LESIONS 1

100010001 T03 TARGET LESIONS 1

100010001 T01 TARGET LESIONS 1

100010001 T02 TARGET LESIONS 2

100010001 T01 TARGET LESIONS 2

100010001 T03 TARGET LESIONS 2

100010002 T02 TARGET LESIONS 0

100010002 T01 TARGET LESIONS 0

100010002 T02 TARGET LESIONS 1

100010002 T01 TARGET LESIONS 1

100010002 T02 TARGET LESIONS 2

100010002 T01 TARGET LESIONS 2

100030002 T03 TARGET LESIONS 0

100030002 T04 TARGET LESIONS 0

100030002 T02 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 1

100030002 T04 TARGET LESIONS 1

100030002 T03 TARGET LESIONS 1

100030002 T02 TARGET LESIONS 1

160020001 T02 TARGET LESIONS 0

160020001 T01 TARGET LESIONS 0

160020001 T02 TARGET LESIONS 1

Re: query

PGStats — Mon, 29 Dec 2014 23:21:23 GMT

What if you had:

100010002 T02 TARGET LESIONS 0

100010002 T01 TARGET LESIONS 0

100010002 T03 TARGET LESIONS 1

100010002 T01 TARGET LESIONS 1

Would you want to flag that discrepancy?

Re: query

Ksharp — Tue, 30 Dec 2014 07:09:45 GMT

Did you just say the each NUMBER have the same number of obs in each and every subjid group ?

data have;
input subjid : $20. number $ cat & $40. visitnum     ;
cards;
100010001           T02         TARGET LESIONS              0
100010001           T01         TARGET LESIONS              0
100010001           T03         TARGET LESIONS              0
100010001           T02         TARGET LESIONS              1
100010001           T03         TARGET LESIONS              1
100010001           T01         TARGET LESIONS              1
100010001           T02         TARGET LESIONS              2
100010001           T01         TARGET LESIONS              2
100010001           T03         TARGET LESIONS              2
100010002           T02         TARGET LESIONS              0
100010002           T01         TARGET LESIONS              0
100010002           T02         TARGET LESIONS              1
100010002           T01         TARGET LESIONS              1
100010002           T02         TARGET LESIONS              2
100010002           T01         TARGET LESIONS              2
100030002           T03         TARGET LESIONS              0
100030002           T04         TARGET LESIONS              0
100030002           T02         TARGET LESIONS              0
100030002           T01         TARGET LESIONS              0
100030002           T01         TARGET LESIONS              1
100030002           T04         TARGET LESIONS              1
100030002           T03         TARGET LESIONS              1
100030002           T02         TARGET LESIONS              1
160020001           T02         TARGET LESIONS              0
160020001           T01         TARGET LESIONS              0
160020001           T02         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              0
160020002           T01         TARGET LESIONS              0
160020002           T03         TARGET LESIONS              0
160020002           T03         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              2
;
run;
proc sql;
 create table temp as
 select * from
 ( select *,count(*) as n
   from have 
    group by subjid     , visitnum)
 group by subjid
  having range(n) ne 0;

quit;

Xia Keshan

Message was edited by: xia keshan

Re: query

rakeshvvv — Tue, 30 Dec 2014 17:10:06 GMT

Thanks Pg....We consider Visitnum=0 as baseline visit.....So what ever number present in baseline should be present in next occuring visits.....if any number not present in visitnum=0 but present in next visit or vice verse....we consider them as discrepany and we should output those.....

Re: query

rakeshvvv — Tue, 30 Dec 2014 17:14:51 GMT

Thanks Xia.... it necessary that number observation in Visit=0 should be same as the number of obseravtions in subsequemt visit....and should output if there is any difference in obseravtion count....Also it would be great if we can add code where we can output where obeservation count do not match from baseline visit and also if NUMBER variable do not match from baseline visit to next visits......

Re: query

Haikuo — Tue, 30 Dec 2014 17:52:55 GMT

I believe you can also do it using Array(), but Hash seems handy:

data have;

input subjid : $20. number $ cat & $40. visitnum;

cards;

100010001 T02 TARGET LESIONS 0

100010001 T01 TARGET LESIONS 0

100010001 T03 TARGET LESIONS 0

100010001 T02 TARGET LESIONS 1

100010001 T03 TARGET LESIONS 1

100010001 T01 TARGET LESIONS 1

100010001 T02 TARGET LESIONS 2

100010001 T01 TARGET LESIONS 2

100010001 T03 TARGET LESIONS 2

100010002 T02 TARGET LESIONS 0

100010002 T01 TARGET LESIONS 0

100010002 T02 TARGET LESIONS 1

100010002 T01 TARGET LESIONS 1

100010002 T02 TARGET LESIONS 2

100010002 T01 TARGET LESIONS 2

100030002 T03 TARGET LESIONS 0

100030002 T04 TARGET LESIONS 0

100030002 T02 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 1

100030002 T04 TARGET LESIONS 1

100030002 T03 TARGET LESIONS 1

100030002 T02 TARGET LESIONS 1

160020001 T02 TARGET LESIONS 0

160020001 T01 TARGET LESIONS 0

160020001 T02 TARGET LESIONS 1

;

run;

data want;

if _n_=1 then

do;

declare hash h1();

h1.definekey('number');

h1.definedone();

declare hash h2();

h2.definekey('number');

h2.definedone();

end;

do until(last.subjid);

set have;

by subjid;

if visitnum=0 then

rc=h1.replace();

else rc=h2.replace();

end;

rc=h1.equals(hash:

'h2', result: flag);

put flag=;

do until (last.subjid);

set have;

by subjid;

if flag ne 1 then

output;

end;

rc=h1.clear();

rc=h2.clear();

drop rc flag;

run;

Re: query

PGStats — Tue, 30 Dec 2014 20:08:09 GMT

To get a detailed list of discrepansies:

data visit;

length subjId $12 cat $16;

input subjid $ number $ cat $& visitnum;

datalines;

100010001 T02 TARGET LESIONS 0

100010001 T01 TARGET LESIONS 0

100010001 T03 TARGET LESIONS 0

100010001 T02 TARGET LESIONS 1

100010001 T03 TARGET LESIONS 1

100010001 T01 TARGET LESIONS 1

100010001 T02 TARGET LESIONS 2

100010001 T01 TARGET LESIONS 2

100010001 T03 TARGET LESIONS 2

100010002 T02 TARGET LESIONS 0

100010002 T01 TARGET LESIONS 0

100010002 T02 TARGET LESIONS 1

100010002 T01 TARGET LESIONS 1

100010002 T02 TARGET LESIONS 2

100010002 T01 TARGET LESIONS 2

100010002 T03 TARGET LESIONS 2

100030002 T03 TARGET LESIONS 0

100030002 T04 TARGET LESIONS 0

100030002 T02 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 1

100030002 T04 TARGET LESIONS 1

100030002 T03 TARGET LESIONS 1

100030002 T02 TARGET LESIONS 1

160020001 T02 TARGET LESIONS 0

160020001 T01 TARGET LESIONS 0

160020001 T02 TARGET LESIONS 1

;

proc sql;

create table visit0 as

select unique a.subjId, b.visitNum, a.number

from visit as a inner join visit as b

on a.subjid = b.subjid

where a.visitnum = 0;

create table newNumbers as

select subjId, visitNum, number

from visit except corr

select * from visit0;

create table missingNumbers as

select subjId, visitNum, number

from visit0 except corr

select * from visit;

title "Discrepansies";

select "New number", * from newNumbers

union all

select "Missing number", * from missingNumbers

order by subjid, visitNum, number;

quit;

Re: query

Ksharp — Wed, 31 Dec 2014 08:18:16 GMT

It is more clear .

data have;
input subjid : $20. number $ cat & $40. visitnum     ;
cards;
100010001           T02         TARGET LESIONS              0
100010001           T01         TARGET LESIONS              0
100010001           T03         TARGET LESIONS              0
100010001           T02         TARGET LESIONS              1
100010001           T03         TARGET LESIONS              1
100010001           T01         TARGET LESIONS              1
100010001           T02         TARGET LESIONS              2
100010001           T01         TARGET LESIONS              2
100010001           T03         TARGET LESIONS              2
100010002           T02         TARGET LESIONS              0
100010002           T01         TARGET LESIONS              0
100010002           T02         TARGET LESIONS              1
100010002           T01         TARGET LESIONS              1
100010002           T02         TARGET LESIONS              2
100010002           T01         TARGET LESIONS              2
100030002           T03         TARGET LESIONS              0
100030002           T04         TARGET LESIONS              0
100030002           T02         TARGET LESIONS              0
100030002           T01         TARGET LESIONS              0
100030002           T01         TARGET LESIONS              1
100030002           T04         TARGET LESIONS              1
100030002           T03         TARGET LESIONS              1
100030002           T02         TARGET LESIONS              1
160020001           T02         TARGET LESIONS              0
160020001           T01         TARGET LESIONS              0
160020001           T02         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              0
160020002           T01         TARGET LESIONS              0
160020002           T03         TARGET LESIONS              0
160020002           T03         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              1
160020002           T02         TARGET LESIONS              2
;
run;
data want(drop=n group basic);
group=0;
do until(last.subjid);
 set have;
 by subjid;

 group+1;n=0;
 do until(last.visitnum);
  set have ;
  by subjid visitnum notsorted;
  n+1;
 end;
 if group=1 then basic=n;
 do until(last.visitnum);
  set have ;
  by subjid visitnum notsorted;
  if group ne 1 and n ne basic then output;
 end;
 
end;
run;

Xia Keshan

Re: query

naveen_srini — Wed, 31 Dec 2014 12:52:01 GMT

I am novice, unlike the greats PG, Xia and Haikuo, hence the long and ugly way, for learning purpse it helps:

data have;/*Raw data copied from Haikuo*/

input subjid : $20. number $ cat & $40. visitnum;

cards;

100010001 T02 TARGET LESIONS 0

100010001 T01 TARGET LESIONS 0

100010001 T03 TARGET LESIONS 0

100010001 T02 TARGET LESIONS 1

100010001 T03 TARGET LESIONS 1

100010001 T01 TARGET LESIONS 1

100010001 T02 TARGET LESIONS 2

100010001 T01 TARGET LESIONS 2

100010001 T03 TARGET LESIONS 2

100010002 T02 TARGET LESIONS 0

100010002 T01 TARGET LESIONS 0

100010002 T02 TARGET LESIONS 1

100010002 T01 TARGET LESIONS 1

100010002 T02 TARGET LESIONS 2

100010002 T01 TARGET LESIONS 2

100030002 T03 TARGET LESIONS 0

100030002 T04 TARGET LESIONS 0

100030002 T02 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 0

100030002 T01 TARGET LESIONS 1

100030002 T04 TARGET LESIONS 1

100030002 T03 TARGET LESIONS 1

100030002 T02 TARGET LESIONS 1

160020001 T02 TARGET LESIONS 0

160020001 T01 TARGET LESIONS 0

160020001 T02 TARGET LESIONS 1

;

run;

data want;
set have;
by subjid visitnum notsorted;
array n(30) _temporary_; *array subscript is arbitrary;
if first.subjid then call missing(of n{*});
if first.visitnum then count=0;
count+1;
if last.visitnum then n(_n_)=count;
if last.subjid then do;
if max(of n

) ne count then discrepancy="yes";
else discrepancy="no";
end;
if discrepancy="yes" then output;
drop count;
run;

proc sort data= have out=have1;
by subjid;
run;

data final;
merge want have1;
by subjid;
if discrepancy="yes" then output;
drop discrepancy;
run;

Regards,

Naveen Srinivasan

L&T Infotech