my brain hurts!

[quietann]

just got through a gnarly data-selection and sorting problem... I had two data sets from the same people. Data set 1 had one observation per person, while dataset 2 had anywhere from one to three observations. I was trying to retain the correct observation (the one with all the data) in a new dataset and get rid of duplicates... but SAS is dumb (or I am dumb, or both) and it kept getting screwed up.

Finally I realized that I needed to deal with the people who had duplicates separately. So here are the steps:

1. Merge the two original datasets, but the merged data go to one temporary dataset if there is a single observation for an ID, but another temporary dataset if there are >1 observations per ID. (Also get rid of any observations for people who are only in the second original data set.)
   
   
2. In the dataset with >1 observation, retain those where some index variable matches for both original datasets. (in this case, "visit=dietvisit")
   
   
3. Append the new new dataset with the single-observation temporary dataset.
   
   Code:
   
   

   /* now add dietary iron values where they are available */
   data new dups;
   	merge fefin (in=g) diet;
   	by alive_id;
   	if not g then delete;		/* delete people who 
                                           only have diet data */
   	if first.alive_id=last.alive_id then output new;	
   
   	else output dups;		/* this is where people 
                                           w/ >1 observation go */
   run;
   
   /* if >1 visit then we want the one where dietary iron was collected */
   
   data nodups;
   	set dups;
   	if (visit ^= '' & dietvisit ^='' & visit=dietvisit);		
   run;
   
   /* append the correct duplicate observations to the new data set */
   
   proc append base=new data=nodups;
   run;