如何按组最后选择SAS中的某些内容

Question

Say I have a table Tbl that is sorted by 3 columns {a,b,c} I also have another 100 columns, one of them is d . How can I flag the last row by a group such that d=something , the flag shall be a new column. Hopefully this is doable withOUT re-sorting the whole table

a b c ...many columns... d IDX
1                        5 1                        
1                        3 2
1                        3 3
2                        3 4 
2                        3 5
2                        2 6
2                        2 7

On this table we want to add another column newCol to flag the last row by group a where d = 3

a b c ...many columns... d IDX newCol
1                        5 1   0                     
1                        3 2   0
1                        3 3   1
2                        3 4   0
2                        3 5   1
2                        2 6   0
2                        2 7   0

Answer 1

data want;
set have;
by a d notsorted;
if last.d and d=3 then flag=1;
run;

This requires the dataset to be sorted in a useful fashion - it doesn't have to be in order by d, but it does have to have all the d's of one value together (ie, not 3 3 1 3 4 1 2 3 but 3 3 3 3 4 1 1 2 is fine).

If that's not the case, then there isn't a solution that doesn't rely on sorting in some fashion, whether it be SQL (which does sort the data, it just doesn't tell you it's doing it), PROC SORT , or a hash table (which if you can fit everything into memory might be the fastest sort).

Answer 2

I'm not sure how this gets implemented, but the following does the work that you want:

proc sql;
    select a, b, c, . . .
    from t
    group by a, b
    having c = max(c);

Note that this syntax is quite specific to SAS proc sql. It is not ANSI standard and will not work in most other databases.

This uses a process called "remerging". I'm not sure if it resorts the original table.

EDIT:

Flagging the lines is just as easy:

proc sql;
    select a, b, c, (case when c = max(c) then 'Y' else 'N' end) as flag, . . .
    from t
    group by a, b;

However, if the data is already sorted, it is probably more efficient to use a data step for this purpose.