PROC MERGE

[Gladis6680]

Hi,

question on merge. I would like to merge 2 data sets

DATA_1
	HOME_a	DESCRIPTION	VEDNOR_a	BRAND_a	DISTRIBUTOR_a	SALES_a	QUANTITY_a
	HOME 1	MUFFINS	VENDOR 1	BRAND 1	DIST 1	$519	10
	HOME 2	MUFFINS	VENDOR 1	BRAND 1	DIST 2	$536	10
TOTAL						$1,055	20

DATA_2
HOME_b	DESCRIPTION	VEDNOR_b	BRAND_b	DISTRIBUTOR	QUANTITY_b	SALES_b	REBATES
HOME 2	MUFFINS	VENDOR 1	BRAND 1	DIST 2	10	$536	$17

the code I wrote is

data COMB;

merge data_1 (in=a)

           data_2(in=b);

by Description

if a=1 and b=1;

run;

I am getting 

HOME_a

DESCRIPTION

VEDNOR_a

BRAND_a

DISTRIBUTOR_a

SALES_a

QUANTITY_a

HOME_b

DESCRIPTION_b

VEDNOR_b

BRAND_b

QUANTITY_b

SALES_b

REBATES

HOME 1

MUFFINS

VENDOR 1

BRAND 1

DIST 1

$519

10

HOME 2

MUFFINS

VENDOR 1

BRAND 1

10

$536

$17

HOME 2

MUFFINS

VENDOR 1

BRAND 1

DIST 2

$536

10

HOME 2

MUFFINS

VENDOR 1

BRAND 1

10

$536

$17

 What I want is 

HOME	DESCRIPTION	VEDNOR	BRAND	DISTRIBUTOR	SALES	QUANTITY	REBATES
HOME 2	MUFFINS	VENDOR 1	BRAND 1	DIST 2	$536	10	$17

I am not sure why the rebates and sales are getting duplicated.

Thanks,

[SASKiwi]

That's because DESCRIPTION contains duplicate values. You need to merge by other variables as well like HOME_2.

[Gladis6680]

Ok, so it looks like i need to have something unique. I am working with transactional data, therefore I find it a little challenging 

DATA 1 has the transactional data with sales only and DATA 2 has transactional data with only rebates. Maybe I need to create a new variable and combine 2 variables to make it unique?

[Tom]

You probably want to INTERLEAVE the observations instead of MERGING them.

Use SET instead of merge and you will get all of the observations from both datasets.

You can then apply your rules to combine them in a way that makes sense.

For example if you wanted to find the total cost after removing the rebates you might do something like to add the COST and subtract the REBATE into a new variable TOTAL and collapse to just one observation per ID group.

data want;
   set costs rebates;
   by id;
   if first.id then total=0;
   total+cost-rebate;
   if last.id then output;
run;
   

[Patrick]

Looking at your sample data it appears the key for merging/joining needs to be:

HOME

DESCRIPTION

VEDNO

BRAND

DISTRIBUTOR