SAS codes for bootstrapping for propensity score matched sample

[schatterjee]

Hello members,

I am trying to implement bootstrapping for generating propensity scores, perform PS matching (GREEDY 5→1 algorithm) and then obtain bootstrapped estimates from the final Cox regression. The final step is not executed and I am receiving the following error message:

"ERROR: Data set PERM.BOOTSAMPLEPSMATCH1 is not sorted in ascending sequence. The current BY group has Sample Replicate Number = 98 and the next BY group has Sample Replicate Number = 3."

Please see my editor below. Your help on this will be highly appreciated! 

%let rep = 100;

proc surveyselect data= cohort out=BOOTSAMPLE

     seed = 1347 method = urs

      samprate = 1 outhits rep = &rep;

run;

PROC LOGISTIC DATA=BOOTSAMPLE DESCENDING;
BY REPLICATE;

CLASS EXPOSURE AGE SEX CATVAR1 CATVAR2 CATVARN /PARAM=REF REF=FIRST;

MODEL NON_SEL = CATVAR1 CATVAR2 CATVARN CONTVAR1;

OUTPUT OUT= BOOTSAMPLEPS(DROP=_LEVEL_) PRED=PROB XBETA=LOGIT;
RUN;

/*============================================================*/
/*PS MATCHING*/
/*============================================================*/

%MACRO GREEDMTCH (Lib=, Dataset=, depend=, matches= );

%macro sortcc;

proc sort data = tctrl out = &lib..scontrol;
  by prob randnum;
run;

proc sort data = tcases out = &lib..scase;
   by prob ;
run;

%mend sortcc;

%macro initcc(digits);

data tcases(drop=cprob)
     tctrl(drop=aprob);
   set &lib..&dataset.;
   if &depend. = 0 then do;
     cprob = round(prob,&digits.);
     cmatch = 0;
     randnum = ranuni(1234567);
     output tctrl;
   end;
   else if &depend. = 1 then do;
      cmatch = 0;
      aprob = round(prob,&digits.);
      output tcases;
   end;
run;

%sortcc;
%mend initcc;

%macro match(matched,digits);

data &lib..&matched.(drop = cmatch randnum aprob cprob start oldi
             curctr1 matched);
    set &lib..scase;
    curob + 1;
    matchto = curob;
    if curob = 1 then do;
       start = 1;
       oldi  = 1;
    end;
    do i = start to n;
        set &lib..scontrol point = i nobs = n;
        if i > n then goto startovr;
        if _error_ = 1 then abort;
        curctr1 = i;
           if aprob = cprob then do;
              cmatch = 1;
              output &lib..&matched.;
              matched = curctr1;
              goto found;
           end;
           else if cprob > aprob then
                goto nextcase;
       startovr:
         if i > n then goto nextcase;
   end;
   nextcase:
       if cmatch = 0 then start = oldi;
   found:
       if cmatch = 1 then do;
          oldi  = matched + 1;
          start = matched + 1;
          set &lib..scase point = curob;
          output &lib..&matched.;
       end;
  retain oldi start;
  if _Error_=1 then _Error_=0;
run;

proc sort data = &lib..scase out = sumcase;
   by ID;
run;

proc sort data = &lib..scontrol out = sumcontrol;
   by ID;
run;

proc sort data = &lib..&matched. out = smatched(keep=BENE_ID matchto);
   by ID;
run;

data tcases(drop=matchto);
   merge sumcase(in = in1)
         smatched;
   by ID;
   if in1 and matchto = .;
   cmatch = 0;
   aprob = round(prob,&digits.);
run;

data tctrl(drop=matchto);
     merge sumcontrol(in = in1)
           smatched;
     by ID;
     if in1 and matchto = .;
     cmatch = 0;
     cprob = round(prob,&digits.);
run;

%sortcc;
%mend match;

%initcc(.000001);

%match(match5,.00001);
%match(match4,.0001);
%match(match3,.001);
%match(match2,.01);
%match(match1,.1);

data &lib..&matches.;
     set &lib..match5(in = a)
         &lib..match4(in = b)
         &lib..match3(in = c)
         &lib..match2(in = d)
         &lib..match1(in = e);
	 if a then matchto = matchto + 1000;
     if b then matchto = matchto + 100000;
     if c then matchto = matchto + 10000000;
     if d then matchto = matchto + 1000000000;
     if e then matchto = matchto + 100000000000;
run;

proc sort data = &lib..&matches. out = &lib..s&matches.;
     by &depend.;
run;

%mend greedmtch;
%GREEDMTCH( Lib=PERM, Dataset=BOOTSAMPLEPS, depend=EXPOSURE, matches=BOOTSAMPLEPSMATCH);


PROC SORT DATA=PERM.BOOTSAMPLEPSMATCH OUT=PERM.BOOTSAMPLEPSMATCH1; 
BY ID REPLICATE; 
RUN;

PROC PHREG DATA=PERM.BOOTSAMPLEPSMATCH1;
BY REPLICATE;
CLASS EXPOSURE /PARAM=REF REF=FIRST;
MODEL  TIMETOEVENT * CENSOR(0) = EXPOSURE /RL;
STRATA MATCHTO;
TITLE "PROPENSITY-SCORE MATCHED BOOTSTRAPPED COX MODEL";
RUN;  

[Reeza]

PROC PSMATCH will do greedy matching for you.

[schatterjee]

Thanks Reeza! Can you suggest a way to resolve the error on bootstrapping in the final step on regression?

[Reeza]

Sorry, I have no idea what the error is. I can't run your code (no data), there are no comments to let me know what you thought was happening and you didn't include the log or error.

[Reeza]

Sorry, you did include the error but didn't indicate which step generates it.

[schatterjee]

Hi Reeza,

I received the error at the proc phreg step.

Thanks,

Satabdi

[Reeza]

Why did you sort by ID? Is that needed? The error is because the BY statement is different between the sort and the PROC PHREG.

[schatterjee]

Thanks so much Reeza! Let me try without id and see what I get.

[schatterjee]

Hi Reeza,

I redid the analyses, and that error was resolved. This time I received a warning "Convergence was not attained in 25 iterations". Also for the final output, all the fields (parameter estimate, std error, 95% CI, Chi-Sq, Pr>Chi-sq, HR, 95% CI of HR) are blank. Is it because convergence was not attained?

Satabdi

[Reeza · Accepted solution]

---

@schatterjee wrote:

Hi Reeza,

I redid the analyses, and that error was resolved. This time I received a warning "Convergence was not attained in 25 iterations". Also for the final output, all the fields (parameter estimate, std error, 95% CI, Chi-Sq, Pr>Chi-sq, HR, 95% CI of HR) are blank. Is it because convergence was not attained?

Satabdi

---

Yes, but that's likely due to a data issue. Check that you have enough data, and that you don't have 0 cells in your categorical data.

[schatterjee]

Ok thank you so much for your help!

[ballardw]

---

@schatterjee wrote:

Hi Reeza,

I received the error at the proc phreg step.

Thanks,

Satabdi

---

Fairly clear error:

ERROR: Data set PERM.BOOTSAMPLEPSMATCH1 is not sorted in ascending sequence.

In

PROC PHREG DATA=PERM.BOOTSAMPLEPSMATCH1;
BY REPLICATE;<= NOT sorted properly here
CLASS EXPOSURE /PARAM=REF REF=FIRST;
MODEL  TIMETOEVENT * CENSOR(0) = EXPOSURE /RL;
STRATA MATCHTO;
TITLE "PROPENSITY-SCORE MATCHED BOOTSTRAPPED COX MODEL";
RUN;  

So sort the data set BY REPLICATE prior to the Proc Phreg.

[schatterjee]

Thanks!