Question

So I have the following SQL code which calculates the pearson correlation between two users rating:

 select  @u1avg:=avg(user1_rating), 
    @u2avg:=avg(user2_rating), 
    @u1sd:=stddev(user1_rating),
    @u2sd:=stddev(user2_rating)
    from
(select r1.userId as User1_id,r1.rating as User1_rating,
        r2.userId as User2_id,r2.rating as User2_rating

from mydb.ratings r1 join mydb.ratings r2 on r1.itemId = r2.itemid 
where r1.userId=1 and r2.userId=2) sample;


select (1/(count(r1.rating-1)))*sum(((r1.rating-@u1avg)/@u1sd)*((r2.rating-@u2avg)/@u2sd))*(count(r1.rating)/(1+count(r1.rating)))

from mydb.ratings r1 join mydb.ratings r2 on r1.itemId = r2.itemid 
where r1.userId=1 and r2.userId=2;

I would like to turn that into a function, for example corr(A,B)
any help would be useful.

The problem I get is that where it says sample saying not allowed or something like that, however if I remove the sample I get an error saying every table must have alias.

Answer 1

I think you can do away with the derived table in the first query which will see off that particular error –

SELECT
    @u1avg:=avg(r1.rating),
    @u2avg:=avg(r2.rating),
    @u1sd:=stddev(r1.rating),
    @u2sd:=stddev(r2.rating)
FROM mydb.ratings r1
INNER JOIN mydb.ratings r2
    ON r1.itemId = r2.itemId
WHERE r1.userId=1
AND r2.userId=2;

SELECT (1/(COUNT(r1.rating-1)))*SUM(((r1.rating-@u1avg)/@u1sd)*((r2.rating-@u2avg)/@u2sd))*(COUNT(r1.rating)/(1+COUNT(r1.rating)))
FROM mydb.ratings r1
INNER JOIN mydb.ratings r2
    ON r1.itemId = r2.itemid
WHERE r1.userId=1
AND r2.userId=2;