Merging two datasets by ID and date

[jwestra84]

Hello,

I am having a little trouble merging two datasets together. 

Dataset A:

ID Date x1 x2 x3

Dataset B:

ID Date y1

Both datasets are one record per date, but A will have 5 dates per ID, whereas B will have anywhere from 0 to 6 dates per ID. Not every ID in A will have records in B.

When I first tried this, it was merging correctly for the first occurance per id, but afterwards would just add the records from B as if I were interleaving the datasets. I used the following code:

data AB; merge A B;

by ID date;

run;

The resulting dataset looks like this:

ID Date x1 x2 x3 y1

1   1      1   0   1

1   2      0   0   1   1

1   3      0   1   1   

1   4      1   1   0

1   5      0   1   1

1   5                     2

So set A has all five dates and set B has records at dates 2 and 5. I would like it so that the two records with ID 1 and date 5 were the same record.

Any help would be appreciated!

Thanks!

[Reeza]

Many to many merges aren't what you expect in a data step. Using SQL in this case is recommended.

You show what you're getting, but not what you expect ...what do you expect as results?

http://support.sas.com/kb/24/752.html

[jwestra84]

I am expecting to get this:

ID Date x1 x2 x3 y1

1   1      1   0   1

1   2      0   0   1   1

1   3      0   1   1   

1   4      1   1   0

1   5      0   1   1   2

There should only be one record per date, but the data step results in there being an extra record for the dates that are also in B.

I have also tried in SQL, but have gotten similar results.

proc sql;
create table AB as
select *
from A as a left join B as b
on a.ID=b.cID and a.date=b.date;
quit;

[PGStats · Accepted solution]

I suspect that your dates don't really match all the time. Try rounding the dates before joining :

proc sql;
create table AB as
select a.*, b.y
from A as a left join B as b
on a.ID=b.ID and round(a.date)=round(b.date);
quit;

[jwestra84]

Thank you for calling my attention to this. Turns out this was the problem.

Thank you!

[AnnaBrown]

I'm glad you were able to identify the problem, jwestra84! If one of the replies was the exact solution to your problem, can you "Accept it as a solution"? Or if one was particularly helpful, feel free to "Like" it. This will help other community members who may run into the same issue know what worked.

Thanks!
Anna

[Astounding]

You will need to check your data.  The results you describe do not match the program you showed.  The program would perform exactly as you hoped and expected.  These results correspond to data set B having 2 observations for DATE=5.  On the first of those, Y1 is missing and on the second, Y1 is 2.