topic Re: Merging - many to many relationship using data step in SAS Programming

Merging - many to many relationship using data step

Sheeba — Tue, 06 Sep 2016 22:33:57 GMT

Hi,

I have two datasets A and B which I am trying to join and the scenario is a many to many join.

I am confused on the join. Does each record of dataset A gets the matching last record from B? Is it possible to have a control here so that each record in dataset A is combined with all the matching records from B

Any help will be greatly appreciated.

Thanks in advance

data A;

input code $ x y;

datalines;

AA 1 2

AA 3 4

AA 5 6

BB 7 8

CC 9 10

;

run;

data B;

input code $ z k;

datalines;

AA 10 20

AA 20 40

BB 70 80

BB 99 77

CC 90 100

;run;

Proc sort data = A;

By code;

Run;

Proc sort data = B;

By code;

Run;

Data c;

Merge A(in =ina) B;

If ina;

Run;

The ooutput i get is

The Actual output which I get is

AA 1 2 10 20

AA 3 4 20 40

AA 5 6 20 40

BB 7 8 70 80

BB 7 8 99 77

CC 9 10 90 100

Regards,

Sheeba Swaminathan

Re: Merging - many to many relationship using data step

ballardw — Tue, 06 Sep 2016 22:42:50 GMT

From the online help:

Note: The MERGE statement does not produce a Cartesian product on a many-to-many match-merge. Instead it performs a one-to-one merge while there are observations in the BY group in at least one data set. When all observations in the BY group have been read from one data set and there are still more observations in another data set, SAS performs a one-to-many merge until all observations have been read for the BY group.

So in your case for the Code=AA. First record of set a matched first of B, then second of A with second of B. B ran out of AA codes so remaing AA in set A matched against the last AA in dataset B.

Many to many joins are usually better left to Proc Sql.

Re: Merging - many to many relationship using data step

Sheeba — Wed, 07 Sep 2016 03:16:57 GMT

HI Ballardw,

Thanks a lot for the details.

Also i assume many to many merge differ from proc sql left join? is that correct.

Thanks again

Regards,

Sheeba Swaminathan

Re: Merging - many to many relationship using data step

LinusH — Wed, 07 Sep 2016 04:33:18 GMT

Yes.
Even if you can, M-M merge in the data step is almost never a valid use case.

Re: Merging - many to many relationship using data step

Sheeba — Wed, 07 Sep 2016 15:42:58 GMT

Hi LinusH,

Thank you.

Regards,

Sheeba Swaminathan