Reputation: 1608
I want to make a query where I am computing the difference between two columns. Something like:
SELECT a,
b,
a - b as "diff"
FROM ...
Now I would like to calculate the stddev of the "diff" column using postgresql built-in stddev aggregate function. How can I achieve this?
Thanks.
EDIT:
The actual query is this:
SELECT tr.date_start,
tr.date_end,
(((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tr.amt_won + tr.cnt_bounty * tourney_summary.amt_bounty) ELSE 0.0 END))) AS "amt_won_curr_conv",
(((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tourney_summary.amt_buyin + tourney_summary.amt_fee + tourney_summary.amt_rebuy * tr.cnt_rebuy + tourney_summary.amt_addon * tr.cnt_addon + tourney_summary.amt_bounty) ELSE 0.0 END))) AS "amt_buyin_ttl_curr_conv",
((((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tr.amt_won + tr.cnt_bounty * tourney_summary.amt_bounty) ELSE 0.0 END))) - (((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tourney_summary.amt_buyin + tourney_summary.amt_fee + tourney_summary.amt_rebuy * tr.cnt_rebuy + tourney_summary.amt_addon * tr.cnt_addon + tourney_summary.amt_bounty) ELSE 0.0 END)))) as net_amt_won,
stddev((((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tr.amt_won + tr.cnt_bounty * tourney_summary.amt_bounty) ELSE 0.0 END))) - (((CASE when(tourney_summary.val_curr_conv != 0) THEN tourney_summary.val_curr_conv * (tourney_summary.amt_buyin + tourney_summary.amt_fee + tourney_summary.amt_rebuy * tr.cnt_rebuy + tourney_summary.amt_addon * tr.cnt_addon + tourney_summary.amt_bounty) ELSE 0.0 END)))) as diff_std_dev
FROM tourney_summary,
tourney_results tr
WHERE
tr.id_player=1
AND tourney_summary.id_tourney = tr.id_tourney
AND ((tourney_summary.id_gametype = 1)
AND (((((((tourney_summary.id_table_type IN
(SELECT lttt.id_table_type
FROM tourney_table_type lttt
WHERE lttt.val_seats = 2))))))
AND (((((tourney_summary.id_table_type IN
(SELECT lttt.id_table_type
FROM tourney_table_type lttt
WHERE position('S' IN lttt.val_speed) > 0))
OR (tourney_summary.id_table_type IN
(SELECT lttt.id_table_type
FROM tourney_table_type lttt
WHERE position('H' IN lttt.val_speed) > 0))))))))
AND ((tourney_summary.date_start >= '2013/08/15 23:00:00')))
GROUP BY tr.date_start,
tr.date_end,
tourney_summary.val_curr_conv,
tr.amt_won,
tr.cnt_bounty,
tourney_summary.amt_bounty,
tourney_summary.amt_buyin,
tourney_summary.amt_fee,
tourney_summary.amt_rebuy,
tr.cnt_rebuy,
tourney_summary.amt_addon,
tr.cnt_addon
ORDER BY tr.date_end DESC;
The "a" and "b" expressions (the ones with CASE) are big. And I don't know how to avoid the copy/paste. In any case using stddev on the a-b expression returns a blank column. What am I doing wrong?
Thanks.
Upvotes: 1
Views: 616
Reputation: 117370
it's also possible to do this with cte
with cte1 as (
select a, b, a - b as diff
from ...
)
select
a, b, diff, stddev(diff) as diff_stddev
from cte1
Upvotes: 2
Reputation: 324415
You pretty much answer it yourself. Calculate the standard deviation of the difference:
SELECT a,
b,
a - b as "diff",
stddev(a - b) AS "diff_stddev"
FROM ...
If a - b
is a computationally expensive operation or is in fact a much more complex expression in reality, you can wrap it in a subquery:
SELECT a, b, "diff", stddev("diff") AS diff_stddev
FROM (
SELECT a, b, a - b
FROM ...
) x (a, b, "diff")
x
is just a throw-away alias for the subquery table.
Upvotes: 3