group by clause in SUBQUERIES in proc sql

[2ratika7]

proc sql;
create table public_school as
select ID,time_period,score,class from data 
where ID = (select ID from data
		where time_period = (select max(time_period )from data))
group by class  
order by time_period 
;quit;

ID	time_period	section	percentage
281	0.25	a_section	83
281	0.5	a_section	82
281	0.75	a_section	98
281	1	a_section	88
281	1.25	a_section	81
282	0.25	a_section	80
282	0.5	a_section	82
282	0.75	a_section	84
282	1	a_section	86
282	1.06	a_section	54
282	1.75	a_section	80
282	2	a_section	82
282	2.25	a_section	84
282	2.5	a_section	86
282	2.75	a_section	88
282	3.0	a_section	90
282	3.25	a_section	92
301	0.25	b_section	70
301	0.5	b_section	76
301	0.75	b_section	82
301	1	b_section	88
302	0.25	b_section	90
302	0.75	b_section	100
303	0.25	b_section	84

In Sample table shown above, I have different no. of students in each section(a,b,c,d) and school_type( public, private) and their final % obtained for each quarterly held exam. Objective is to select student ID with maximum number of time period from each section. Output table should should look like(including score also) :

ID	time_period	section
282	0.25	a_section
282	0.5	a_section
282	0.75	a_section
282	1	a_section
282	1.25	a_section
282	1.5	a_section
282	1.75	a_section
282	2	a_section
282	2.25	a_section
282	2.5	a_section
282	2.75	a_section
301	0.25	b_section
301	0.5	b_section
301	0.75	b_section
301	1	b_section
:	:	:

I have tried code displayed above , but getting error as:

A GROUP BY clause has been discarded because neither the SELECT clause nor the optional HAVING clause of the associated table-expression referenced a summary function

Also, if u can tell how to create macro on this code for school type, it will be quite helpful 

[PGStats]

To get all the data from the "student ID with maximum number of time period from each section" :

proc sql;
create table public_school as
select a.*
from 
	data as a inner join
	(select section, id, count(time_period) as total_period 
	 from data 
	 group by section, id) as b on a.section=b.section and a.id=b.id
group by a.section
having total_period = max(total_period)
;
quit;

(untested)

or maybe you'l want to replace count(time_period) with sum(time_period).

[2ratika7]

I have got this note: The query requires remerging summary statistics back with the original data & dataset created with 0 rows. 

[PGStats · Accepted solution]

The remerging message is intended and expected, it's just a feature of SAS/SQL. When I run this test, I get 16 rows into table Public_School:

data have;
input ID  time_period section $ percentage;
datalines;
281 0.25    a_section   83
281 0.5 a_section   82
281 0.75    a_section   98
281 1   a_section   88
281 1.25    a_section   81
282 0.25    a_section   80
282 0.5 a_section   82
282 0.75    a_section   84
282 1   a_section   86
282 1.06    a_section   54
282 1.75    a_section   80
282 2   a_section   82
282 2.25    a_section   84
282 2.5 a_section   86
282 2.75    a_section   88
282 3.0 a_section   90
282 3.25    a_section   92
301 0.25    b_section   70
301 0.5 b_section   76
301 0.75    b_section   82
301 1   b_section   88
302 0.25    b_section   90
302 0.75    b_section   100
303 0.25    b_section   84
;

proc sql;
create table public_school as
select a.*
from 
    have as a inner join
    (select section, id, count(time_period) as total_period 
     from have 
     group by section, id) as b on a.section=b.section and a.id=b.id
group by a.section
having total_period = max(total_period)
;
select * from public_school;
quit;

[2ratika7]

thankyou!