Summarize: Var#1- once per group, Var#2-n - all

[Konkordanz]

Hi forum,

I want to summarize the Account1-AccountN over all regions and per year (2019, 2020).

The population I want to summarize per year, but only one value per Region (for example 2019: 82320980+50879514+43025905). Because otherwise the population size would be too high (the population size of the different regions can be found multiply in the dataset).

Data I have:

year	Region	population	Account1	Account2	AccountN
2019	RegionA	82320980	.	8	7
2019	RegionA	82320980	4	.	8
2019	RegionB	50879514	.	6	.
2019	RegionB	50879514	2	.	.
2019	RegionC	43025905	4	.	5
2019	RegionC	43025905	.	2	.
2020	RegionA	82320520	5	4	.
2020	RegionA	82320520	1	.	8
2020	RegionB	50879320	.	3	5
2020	RegionB	50879320	4	4	.
2020	RegionC	43025782	7	.	6
2020	RegionC	43025782	5	9	.

This is the code for the dataset:

data have;
input year $ Region $ population Account1 Account2 AccountN;
datalines;
2019 RegionA 82320980 . 8 7
2019 RegionA 82320980 4 . 8
2019 RegionB 50879514 . 6 .
2019 RegionB 50879514 2 . .
2019 RegionC 43025905 4 . 5
2019 RegionC 43025905 . 2 .
2020 RegionA 82320520 5 4 .
2020 RegionA 82320520 1 . 8
2020 RegionB 50879320 . 3 5
2020 RegionB 50879320 4 4 .
2020 RegionC 43025782 7 . 6
2020 RegionC 43025782 5 9 .
;

This is what I want:

year	population	Account1	Account2	AccountN
2019	176226399	10	16	20
2020	176225622	22	20	19

I tried it with this code...

proc summary data=have nway;
class year;
id population;
var account1 account2 accountN;
output
out=want (drop=_freq_ _type_)
sum()=
;

...but It doesnt work. It summarizes the account-values per year, but not the population. Where is the problem of my attemption? How can I modify the summary-command? Thank you!

[Kurt_Bremser · Accepted solution]

That's the problems you get with redundant values.

You need a two-step approach: first summarize by year and region, and take the max of population, then summarize by year:

proc summary data=have nway;
class year region;
var population account:;
output
  out=inter
  max(population)=
  sum(acc:)=
;
run;

proc summary data=inter nway;
class year;
var population acc:;
output
  out=want (drop=_:)
  sum()=
;
run;

[Konkordanz]

Thank you, its working. may I ask:

What does that mean? Why an equalsign with no value behind it?

max(population)=
sum(acc:)=

[Kurt_Bremser]

Look under

output-statistic-specification(s)

here:

OUTPUT Statement