SQL Query (incorrect use of joins?) - sql

Question

I'm trying to write an (Oracle) SQL query that, given an "agent_id", would give me a list of questions that agent has answered during an assessment, as well as an average score over all of the times that agent has answered those questions.
Note: I tried to design the query such that it would support multiple employees (so we can query at the store level), hence the "IN" condition in the where clause.
Here's what I have so far:
select question.question_id as questionId,
((sum(answer.answer) / count(answer.answer)) * 100) as avgScore
from SPMADMIN.SPM_QC_ASSESSMENT_ANSWER answer
join SPMADMIN.SPM_QC_QUESTION question
on answer.question_id = question.question_id
join SPMADMIN.SPM_QC_ASSESSMENT assessment
on answer.assessment_id = assessment.assessment_id
join SPMADMIN.SPM_QC_SUB_GROUP_TYPE sub_group
on question.sub_group_type_id = sub_group.sub_group_id
join SPMADMIN.SPM_QC_GROUP_TYPE theGroup
on sub_group.group_id = theGroup.group_id
where question.question_id in (select distinct question2.question_id
from SPMADMIN.SPM_QC_QUESTION question2
)
and question.bool_yn_active_flag = 'Y'
and assessment.agent_id in (?)
and answer.answer is not null
order by theGroup.page_order asc,
sub_group.group_order asc,
question.sub_group_order asc
Basically I would want to see:
|questionId|avgScore|
| 1 | 100 |
| 2 | 50 |
| 3 | 75 |
Such that every question that employee has ever answered is in the list of question indexes with their average score over all of the times they've answered it.
When I run it as is, I'm given a "ORA-00937: not a single-group group function" error. Any sort of combination of a "group by" clause I've added hasn't helped in the least.
When I run it removing the question.question_id as questionId, part of the select, it runs fine, but it shows their average score over all questions. I need it broken down by question.
Any help or pointers would be greatly appreciated.

Answer 1

When you have an aggregate function in the SELECT list (SUM and COUNT are aggregate functions), then any other columns in the SELECT list need to be in a GROUP BY clause. For example:
SELECT fi, COUNT(fo)
FROM fum
GROUP BY fi
The COUNT(fo) expression is an aggregate, the fi column is a non-aggregate. If you were to add another non-aggregate to the SELECT list, it would also need to be included in the GROUP BY. For example
SELECT TRUNC(fee), fi, COUNT(fo)
FROM fum
GROUP BY TRUNC(fee), fi
To be a little more precise, rather than say "columns in the SELECT list", we should actually say "all non-aggregate expressions in the SELECT list" will need to be included in the GROUP BY clause.

Answer 2

It's not your joins but your use of GROUP BY.
When you use a GROUP BY in SQL, the things you GROUP BY are the things which define the groups. Everything else you have in your SELECT have to be in aggregates which operate over the group.
You can also do aggregates over the entire set without a GROUP BY, but then every column will need to be within an aggregate function.