Hi Pierre, It looks like you're saying that each row has an id plus three numeric columns, and you want the stddev calculated from the three numeric columns? In that case you could do this: create table foo (id integer, a float, b float, c float); insert into foo values (1, 2,3,4); insert into foo values (2, 2,3,4); select id, stddev(x) from (select id, unnest(array[a,b,c]) x from foo) bar group by id; id | stddev ----+-------- 1 | 1 2 | 1 (2 rows) But if that's correct, then I think your table is badly structured for a relational database. It might be better to have just two columns: an id and *one* numeric value. Or perhaps an id and an array of numeric values if you really want all values in one row. At a higher level, if you are really taking the stddev of a sample of size 3, you should reconsider applying statistical analysis to your problem at all. I hope this helps! Paul On Wed, Jan 21, 2015 at 10:09 AM, Raymond O'Donnell <rod@xxxxxx> wrote: > On 21/01/2015 18:02, Pierre Hsieh wrote: >> Hi Raymond, >> >> Thanks for your reply. Please see detail as following. Thanks again. > > Can you describe *in words* what sort of calculation you want to do? > > Ray. > > > -- > Raymond O'Donnell :: Galway :: Ireland > rod@xxxxxx > > > -- > Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) > To make changes to your subscription: > http://www.postgresql.org/mailpref/pgsql-general -- _________________________________ Pulchritudo splendor veritatis. -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
