Bayesian ratings in postgresql - database

Question

I have the following table in my db
Name | total_stars | total_reviews
Item A 27 7
Item B 36 9
Item C 27 7
Item D 30 6
Item E 0 0
Item F 0 0
Item F 15 3
I was looking at this article and was trying to implement bayesian rankings in postgresql database.
The formula given for the rank is
br = ( (avg_num_votes * avg_rating) + (this_num_votes * this_rating) ) /
(avg_num_votes + this_num_votes)
where:
avg_num_votes: The average number of votes of all items that have num_votes>0
avg_rating: The average rating of each item (again, of those that have num_votes>0)
this_num_votes: number of votes for this item
this_rating: the rating of this item
This is the query I came up with, but it is not working:
with avg_num_votes as (
select AVG(total_reviews)
from business
where total_reviews != 0),
avg_rating as (
select AVG(total_stars/total_reviews)
from business
where total_reviews != 0)
select * from business
order by ((avg_num_votes * avg_rating) + (total_stars)) / (avg_num_votes + total_reviews);
I am getting: ERROR: column "avg_num_votes" does not exist

Answer 1

Operator WITH in Postgres is only used to create additional working queries to be used in main query.
This query is using sub-selects in FROM clause instead and works as you expect:
SELECT business.* FROM business,
(SELECT avg(total_reviews) AS v
FROM business
WHERE total_reviews != 0
) AS avg_num_votes,
(SELECT avg(total_stars/total_reviews) AS v
FROM business
WHERE total_reviews != 0
) AS avg_rating
ORDER BY ((avg_num_votes.v * avg_rating.v) + (total_stars)) / (avg_num_votes.v + total_reviews)
EDIT: Actually, using WITH is also possible, but does not seem to be shorter compared to first form. Also, it is less portable - first solution will work on MySQL, but this will not:
WITH
avg_num_votes AS (
SELECT avg(total_reviews) AS v
FROM business
WHERE total_reviews != 0
),
avg_rating AS (
SELECT avg(total_stars/total_reviews) AS v
FROM business
WHERE total_reviews != 0
)
SELECT business.*
FROM business, avg_num_votes, avg_rating
ORDER BY ((avg_num_votes.v * avg_rating.v) + (total_stars)) / (avg_num_votes.v + total_reviews)

Answer 2

SQL Fiddle
with av as (
select avg(total_reviews) avg_num_votes
from business
where total_reviews > 0
), ar as (
select name, avg(total_stars * 1.0 / total_reviews) avg_rating
from business
where total_reviews > 0
group by name
)
select b.*, avg_rating, avg_num_votes,
(avg_num_votes * avg_rating + total_stars)
/
(avg_num_votes + total_reviews) br
from
business b
left join
ar on ar.name = b.name
inner join
av on true
order by br, b.name
;
name | total_stars | total_reviews | avg_rating | avg_num_votes | br
--------+-------------+---------------+--------------------+--------------------+------------------------------------
Item A | 27 | 7 | 3.8571428571428571 | 6.4000000000000000 | 3.85714285714285712238805970149254
Item C | 27 | 7 | 3.8571428571428571 | 6.4000000000000000 | 3.85714285714285712238805970149254
Item B | 36 | 9 | 4.0000000000000000 | 6.4000000000000000 | 4.00000000000000000000000000000000
Item D | 30 | 6 | 5.0000000000000000 | 6.4000000000000000 | 5.00000000000000000000000000000000
Item F | 0 | 0 | 5.0000000000000000 | 6.4000000000000000 | 5.00000000000000000000000000000000
Item F | 15 | 3 | 5.0000000000000000 | 6.4000000000000000 | 5.00000000000000000000000000000000
Item E | 0 | 0 | | 6.4000000000000000 |