LOCF

[112211]

data ONE ;
input id visit sbp ;
cards ;
101 10 95
101 20 .
101 30 90
102 10 100
102 20 105
102 30 .
103 10 90
103 20 90
103 30 115
104 10 .
104 20 100
104 30 .
104 40 105
;
proc sort data = one out = sorted ; by id visit ;run;

I want LOCF BOCF WOCF for the above data I tried my code but I didn't get the desired output, Please any anyone help me write code?

Here is the code ;

data lf ;

 set sorted ;

 retain lo;

 by id visit ;

       if first.id  and sbp = . then lo = . ;

          else do ;

             if id ne . then lo = sbp ;

       else sbp = lo ;

  end;

 run;

Need output like below :

LOCF means Last observation carryforward

BOCF means Baseline observation carryforward  (Here baseline visit as 10)

WOCF means Worst observation carryforward

Obs    id     visit    sbp     locf     bocf     wocf
1         1 01      10       95       95          95       95
2        101       20       .           95          95      95
3        101       30       90       90         90      90
4        102       10       100    100         100   100
5         102       20      105     105         105    105
6         102        30        .        105          100   105
7         103        10      90       90          90       90
8         103        20      90       90          90       90
9         103        30       115       115         115      115
10        104        10          .           .             .        105
11         104         20     100       100      100     100
12         104         30       .           100        .         105
13         104           40   105        105       105     105

[Quentin]

Can you show the output data you want?

For simple last obs carry forward, you probably want to use the first value per ID.  And after that, use the value of SBP if SBP is not a missing value.  You can do that like:

data lf ;
  set sorted ;
  retain lo;
  by id visit ;
  if first.id then lo = sbp ;
  else if not missing(sbp) then lo=sbp ;
run ;

proc print ;
run ;

I assume BOCF and WOCF are something like "best" and "worst" ?

I think you could take above and adapt it.  The ELSE statement would compare the value of SBP to the value of LO and then decide whether or not to update the value of LO.

[SASKiwi]

You haven't provided any definitions of what you mean by LOCF, BOCF and WOCF. We need those before we can help code for them.

[mkeintz]

LOCF (presumably "last" observation carried forward) typically means you want to carry forward the most recent non-missing value of a variable

I guess BOCF and WOCF mean "best" and "worst" observation carried forward.  If "best" is highest, and "worst" is lowest, then

dm 'clear log';
data ONE ;
  input id visit sbp ;
cards ;
101 10 95
101 20 .
101 30 90
102 10 100
102 20 105
102 30 .
103 10 90
103 20 90
103 30 115
104 10 .
104 20 100
104 30 .
104 40 105
run;

data want;
  set one;
  by id visit;

  retain locf bocf wocf;
  locf=ifn(first.id,.,coalesce(lag(sbp),locf));
  bocf=ifn(first.id,.,max(bocf,locf));
  wocf=ifn(first.id,.,min(wocf,locf));
run;
proc print;
  id id;
run;

[Tom]

@mkeintz I don't think that logic works.

Those goofy IFN,IFC functions never make much sense to me (ripoffs from EXCEL I think).

If think you are looking for something like:

data want;
  set one;
  by id visit;
  retain locf bocf wocf;
  if first.id then call missing(locf,bocf,wocf);
  locf=coalesce(sbp,locf);
  bocf=max(bocf,locf);
  wocf=min(wocf,locf);
run;

Result

id=101

visit    sbp    locf    bocf    wocf

  10      95     95      95      95
  20       .     95      95      95
  30      90     90      95      90


id=102

visit    sbp    locf    bocf    wocf

  10     100     100     100     100
  20     105     105     105     100
  30       .     105     105     100


id=103

visit    sbp    locf    bocf    wocf

  10      90      90      90     90
  20      90      90      90     90
  30     115     115     115     90


id=104

visit    sbp    locf    bocf    wocf

  10       .       .       .       .
  20     100     100     100     100
  30       .     100     100     100
  40     105     105     105     100

[mkeintz]

---

@Tom wrote:

@mkeintz I don't think that logic works.

Those goofy IFN,IFC functions never make much sense to me (ripoffs from EXCEL I think).

If think you are looking for something like:

data want;
  set one;
  by id visit;
  retain locf bocf wocf;
  if first.id then call missing(locf,bocf,wocf);
  locf=coalesce(sbp,locf);
  bocf=max(bocf,locf);
  wocf=min(wocf,locf);
run;

@Tom 

In this case, I took the OP's request as defining "carry-forward" as excluding the current observation, which I should have clearly stated.  And which is also the reason my results differed from yours.

But even if that assumption is wrong, I'm still a fan of the IFN and IFC functions - for good reason.  They are particularly beneficial for situations such as conditionally retrieving lags.   Since IFN and IFC evaluate both of the alternative return expressions, they always update the lag queue in those expressions, even if the lag is not the returned value.  This allows the replacement of these 4 lines

_tempvar=lag(x);
if some-condition then result=_tempvar;
else result=y;
drop _tempvar;

with this single line

result=ifn(some-condition,lag(x),y);

[Ksharp · Accepted solution]

/*Assuming I understood what you mean.*/
data ONE ;
input id visit sbp ;
cards ;
101 10 95
101 20 .
101 30 90
102 10 100
102 20 105
102 30 .
103 10 90
103 20 90
103 30 115
104 10 .
104 20 100
104 30 .
104 40 105
;

data want;
 do until(last.id);
  set one;
  by id;
  if first.id then baseline=sbp;
  max=max(max,sbp);
 end;
 do until(last.id);
  set one;
  by id;
  if not missing(sbp) then locf=sbp;
  bocf=coalesce(sbp,baseline);
  wocf=coalesce(sbp,max);
  output;
 end;
 drop max baseline;
run;

[allenebrick]

---

@Ksharp wrote:

/*Assuming I understood what you mean.*/
data ONE ;
input id visit sbp ;
cards ;
101 10 95
101 20 .
101 30 90
102 10 100
102 20 105
102 30 .
103 10 90
103 20 90
103 30 115
104 10 .
104 20 100
104 30 .
104 40 105
;

data want;
 do until(last.id);
  set one;
  by id;
  if first.id then baseline=sbp;
  max=max(max,sbp);
 end;
 do until(last.id);
  set one;
  by id;
  if not missing(sbp) then locf=sbp;
  bocf=coalesce(sbp,baseline);
  wocf=coalesce(sbp,max);
  output;
 end;
 drop max baseline;
run;

---

Thankyou, it helped me in understanding my queries.