Question

I’ve got two datasets, one is a subset of the other.

For example let’s say I have

Master table:

Name,status,date
john,born,1-08-2011
frank,alive,1-08-2011
john,alive,1-09-2011
frank,alive,1-09-2011
frank,alive,1-10-2011
john,dead,1-11-2011
frank,alive,1-11-2011

Sub table

frank,alive,1-11-2011
john,dead,1-11-2011

I’d like to search the master table, for each person’s status, on whatever previous day we have a record for.

So my result table I’d like to have

frank,alive,1-10-2011
john,alive,1-09-2011 (since he didn't get a record entry on 1-10)

And then ideally, suppress / remove records where the persons status hasn’t changed.

Answer 1

You can do this in two passes.

The first pass would be to sort your data.

proc sort data=dataset;
    by name date;
run;

Once your dataset is sorted, you can use a data step to iterate through it and use the “first.” and “last.” metavariables.

data second_to_last(drop=last_name last_status last_date);
    retain last_name '' last_status '' last_date .;
    set dataset;
    by name date;
    if first.name then do;
        last_name = '';
        last_status = '';
        last_date = .;
    end;
    if last.name and last_name != '' then do;
        name = last_name;
        status = last_status;
        date = last_date;
        output second_to_last;
    end;
    last_name = name;
    last_status = status;
    last_date = date;
run;

Every time the data step iterates with “first.name” as true, it will clear your retain variables. Right before the end of a group of names, when “last.name” is true, it will set the variables back to what they were previously (if there was a previous row) and output the row.