If value is missing then bring in value from other table (left join dropping rows)

[mhoward2]

Hi! I am having the worst time of my life right now with what should be a super simple procedure. I have three tables. The first table (A) shows all the possible combinations of Group and Bucket. The second table (B) has values for random combinations of Group and Bucket, but also has another field, Type, that comes into play. And the third table (C) has Avg_Values for each combination of Bucket and Type. What I want is the fourth table (D), where it brings in all possible combinations of Group and Bucket for both types, and matches them with their value from table (B). But if (B) has no matching value, then grab the Avg_Value for that Bucket and Type from table (C).

(A)

Group	Bucket
Red	a
Red	b
Red	c
Blue	a
Blue	b
Blue	c
Green	a
Green	b
Green	c

(B)

Type	Group	Bucket	Value
First	Red	a	2
First	Red	b	9
First	Red	c	1
First	Blue	a	3
First	Blue	c	8
First	Green	b	4
First	Green	c	5
Second	Red	a	4
Second	Red	c	7
Second	Blue	a	9
Second	Blue	b	8
Second	Green	a	10
Second	Green	b	2
Second	Green	c	1

(C)

Type	Bucket	Avg_Value
First	a	99
First	b	89
First	c	79
Second	a	69
Second	b	59
Second	c	49

(D)

Type	Group	Bucket	New_Value
First	Red	a	2
First	Red	b	9
First	Red	c	79
First	Blue	a	3
First	Blue	b	89
First	Blue	c	8
First	Green	a	99
First	Green	b	4
First	Green	c	5
Second	Red	a	4
Second	Red	b	59
Second	Red	c	7
Second	Blue	a	9
Second	Blue	b	8
Second	Blue	c	49
Second	Green	a	10
Second	Green	b	2
Second	Green	c	1

It really should be simple but can't for the life of me figure it out. Can anyone here help? Obviously I am in serious distress since its a Saturday before a holiday and I'm working on this lol. Thanks in advance!

[Angel_Larrion · Accepted solution]

The following code does what you want:

data a; /*This will be the base data set (the one that has all the cases)*/
length Type $8 Group $8 Bucket $8;
set a;
type="First";
output;
type="Second";
output;
run;




proc sql;
create table d as
select a.type ,a.group ,a.bucket ,coalesce(b.Value,c.Avg_Value) as New_Value
from a left join b on (a.type=b.type and a.group=b.group and a.bucket=b.bucket)  /*In this join we take the values from b*/
	   left join c on (a.type=c.type and a.bucket=c.bucket)					     /*In this join we take the avg values from c*/
																				/*the coalesce function takes the first non missing value from its arguments*/

order by a.type, a.group desc, a.bucket;                                                                                
quit;

Note: in your example you had

data set b

First

Red

c

1

data set d

First

Red

c

79

 , I assumed it was a mistake.

[mhoward2]

Oh my god you are a beautiful person thank you so much! I knew it was super simple I just couldn't wrap my head around it for some reason. Thank you and I love you.