unable to workout with lag

[vraj1]

pat	par	vis	base	chg	fac
101	let	10	56	2
101	let	10			gre
102	let	10	45	4
102	let	10			gre
103	min	10	44	5
104	min	10	56	8
104	min	10			gre

par

I have the above sample data and want to lag the missing base,chg and fac to the next observation if it is missing and fac =gre.

data inc;
	set samT;

	if vis=10 and  fac= "" then
		do;
			if first.par then
				do;
					chg1=lag1(chg);
					pchg1=lag2(pchg);
				end;
		end;
run;

[RW9]

Do not change the value you are lagging, or base if statements of lag.  If you need to drop down values then a retain is far better, or you can use the update function of datastep.

Retain:

data want (drop=lst_base lst_chg);
  set have;
  retain lst_base lst_chg;
  by pat;
  if first.pat then do;
    lst_base=base;
    lst_chg=chg;
  end;
  else do;
    if base=. then base=lst_base;
    if chg=. then chg=lst_chg;
  end;
run;

**However**, that being said, it looks to me like fac has been merged onto the data somewhere, and has not been merged correctly.  You would not for instance have two rows of the same data for vis=10, one with fac, one without.  This does not make sense, there should be:

pat	par	vis	base	chg	fac
101	let	10	56	2	gre

And only that row for 101, should it not?

[vraj1]

It somehow does'nt work if i use the below code. ot sure where i am doing wrong.

proc sort data=WORK.FILTER_FOR_ADLB_SAS7BDAT_0003 out=inh(keep=usubjid paramcd base chg aval studyid2 visitnum visit);
	by usubjid paramcd visit; run;
data want (drop=lst_base lst_chg);
  set inh;
  retain lst_base lst_chg;
  by paramcd;
  if first.paramcd and visitnum=10 and aval ne . and  studyid2="0A23" then do;
    lst_base=base;
    lst_chg=chg;
  end;
  else do;

data:

     if base=. then base=lst_base;
    if chg=. then chg=lst_chg;
  end;
run;

usubjid	paramcd	aval	base	chg	visit	visitnum	studyid2
101	Bio	45	42	3		.	0A23
101	Bio	43	42	1		.	0A24
101	Bio	45	.	.		.	0A23
101	Bio	43	.	.		.	0A24
101	Bio	43	45	-2		.	0A24
101	Bio	43	42	1	Unscheduled	5.1	0A24
101	Bio	43	.	.	Unscheduled	5.1	0A24
101	Bio	43	45	-2	Unscheduled	5.1	0A24
101	Bio	45	45	.	Visit 1 (Baseline Extension A)	1	0A24
101	Bio	42	42	.	Visit 1 (Screening)	1	0A23
101	Bio	48	42	6	Visit 10 (End of Week 😎	10	0A23
101	Bio	48	.	.	Visit 10 (End of Week 😎	10	0A23
101	Bio	45	42	3	Visit 12 (Completion/Withdrawal)	12	0A23
101	Bio	45	.	.	Visit 12 (Completion/Withdrawal)	12	0A23
101	Bio	45	42	3	Visit 6 (End of Week 4)	6	0A23
101	Bio	45	.	.	Visit 6 (End of Week 4)	6	0A23

[Astounding · Accepted solution]

Here's a quick and easy way ... if it will do what you want.  This will bring forward ALL variables, not just the few that you list.

Assuming your data is sorted by PAT:

data want;

update have (obs=0) have;

by pat;

output;

run;