按条件将多行数据转换为单行

Question

I have data that looks like this (the last column added):

ID      Var 1       Date    What I Want
aa11     Stage I    1980    Delete
aa11     Stage 2    1980    Keep
aa22     Stage 1    1980    Keep
aa22     Stage 2    1990    Delete
aa33     Stage 3    1992    Keep

But I want it to look like this:

ID  Var 1   Date
aa11    Stage 2 1980
aa22    Stage 1 1980
aa33    Stage 3 1992

I want a single row of data per id on these conditions: 1. The entry with the earliest data is taken Else 2. If there are two entries in the same year, take the entry with higher stage (var 1) Else 3. Take the only entry given. How would you go about writing a piece of SQL code or SAS Data-step for this succinctly?

Answer 1

This is a prioritization query. These are tricky. Here is a method using variables to enumerate the rows:

select t.*
from (select t.*,
             (@rn := if(@id = id, @rn + 1,
                        if(@id := id, 1, 1)
                       )
             ) as seqnum
      from t cross join
           (select @rn := 0, @id := '') params
      order by id, year asc, var1 asc
     ) t
where seqnum = 1;

The logic for prioritization is being handled by the order by clause. The rows are enumerated for each id based on the additional keys. The outer query then takes the first row encountered.

Answer 2

I believe that aggregation can be used here without using any variables, however I must make one assumption first: - The Var 1 data can be ordered sequentially ie: Stage 1 < Stage 2 < Stage 3 etc.

If so you could write the following to return what you are looking for:

select
    ID
    --Aggregate results by Max Var1 value
    , max(Var1) as Var1
    , [Date]
from
    [YourTable] a
    --Derived Table to return ID and Var1 by lowest Date
    inner join
        (
            select
                ID
                , Var1
                , min([Date]) as [Date]
            from    
                [YourTable]
            group by
                ID
                , Var1
        ) b on a.ID = b.Id
group by
    Id
    , [Date]

In the event that there is only one value that will be returned for that ID as it has both the MIN Date value and the Max Var1 value.

Answer 3

In SAS this is very simple with a data step. Just sort the data in the required order, then use first.id in a data step to extract the first id. I've assumed that 'Stage I' in your post is a typo and should say 'Stage 1'

/* create original data */
data have;
infile datalines dsd;
input ID $ Var_1 $ Date; 
datalines;
aa11,Stage 1,1980
aa11,Stage 2,1980
aa22,Stage 1,1980
aa22,Stage 2,1990
aa33,Stage 3,1992
;
run;

/* sort dataset */
proc sort data=have;
by id date descending var_1;
run;

/* extract first id only */
data want;
set have;
by id;
if first.id;
run;