Create imputation flag for missing month and day

[westbestern]

I'm working on coding SDTM to ADaM and I want to create an imputation flag "M" when both month and day are missing (ex: 2014 instead of 2014-06-18) and create a new variable called DTHDT (date of death). I have an example below of what I did for a missing day. How do I go about doing this for a missing month and day?

/* if day is missing in dthdtc then do */
if length(dthdtc)=7 then do;
dthdt= input(strip(dthdtc) || '-01', yymmdd10.); *concatenating '-01' to the end of the character 
date string and converting to numeric;
dthdtf='D'; *flag is 'D' when day is imputed;
end;

[Reeza]

/* if day is missing in dthdtc then do */
if length(dthdtc)=7 then do;
         dthdt= input(strip(dthdtc) || '-01', yymmdd10.); 
       *concatenating '-01' to the end of the character date string and converting to numeric;
          dthdtf='D'; 
          *flag is 'D' when day is imputed;
end;

 Similarly?

/* if month&  day is missing in dthdtc then do */
if length(dthdtc)=4 then do;
         dthdt= input(strip(dthdtc) || '07-15', yymmdd10.); 
       *concatenating '07-15' to the end of the character date string and converting to numeric;
          dthdtmnthf='D'; 
          *flag is 'D' when month day is imputed;
end;

Why did you impute day as 1?

Typically for date imputations I've seen the 15th and July used as cutoffs for population/research statistics. It has something to do with allowing it to balance out over time but I suspect it's not too important. 

---

@westbestern wrote:

I'm working on coding SDTM to ADaM and I want to create an imputation flag "M" when both month and day are missing (ex: 2014 instead of 2014-06-18) and create a new variable called DTHDT (date of death). I have an example below of what I did for a missing day. How do I go about doing this for a missing month and day?

/* if day is missing in dthdtc then do */
if length(dthdtc)=7 then do;
dthdt= input(strip(dthdtc) || '-01', yymmdd10.); *concatenating '-01' to the end of the character 
date string and converting to numeric;
dthdtf='D'; *flag is 'D' when day is imputed;
end;

---

[tarheel13]

I think @westbestern meant the flag is dthdtf='M' when month is imputed.

[tarheel13]

what are you imputing month and day to when both are missing? show us the imputation rules. 

[westbestern]

Sorry about that! When month and day are missing, I impute January 1st

[tarheel13]

Okay, that's the way I did it in the code I gave you. Please test it out and feel free to accept as solution if it works for you.

[tarheel13 · Accepted solution]

data have;
input dthdtc :$18.;
datalines;
2014
2014-06
2014-06-18
;
run;

data want;
	set have;
	format dthdt yymmdd10.;
	year=scan(dthdtc,1,'-');
	day=scan(dthdtc,3,'-');
	mth=scan(dthdtc,2,'-');
	if nmiss(year,day,mth)=0 then dthdt=input(catx('-',year,mth,day),yymmdd10.);
	*only day missing;
	if day='' and mth ne '' then do;
		day='01';
		dthdtf='D';
		dthdt=input(catx('-',year,mth,day),yymmdd10.);
	end;
	*day and month both missing;
	if day='' and mth='' then do;
		day='01';
		mth='01';
		dthdtf='M';
		dthdt=input(catx('-',year,mth,day),yymmdd10.);
	end;
run;

you didn't tell us the imputation rules. I imputed day to '01' when it's missing and imputed both day and month to '01' when both are missing.