Help build a query !!!

Question

There are two tables. One consists of members with expenses and next with fixed percentile. I need to build a single query based upon following
1. Need to know the number of members that fall under a fixed range of percent like 10%, 40%, 50%. Suppose, total no of members = 50 then
10% -> 5
40% -> 20
50% -> 25
2. Now I need to rank the members as per their expenses. So, if following be the expenses of the members sorted by expenses
A1 100
A2 99
A3 98
A4 97
A5 96
A6 95
A50 50
then
total expense should be found out as
10% [sum of top 5 : From A1 to A5] = 100 + 99 + 98 + 97 + 96 = 490
40% [sum of next 20: From A6 to A26]
50% [sum of next 25: From A27 to A50]
Please help !!!!!
Edited by: josleen on Jan 11, 2010 7:50 AM

Answer 1

Hi,
There are a number of analytic functions for dealing with percentiles.
Depending on how you want to handle ties, how you count duplicate values, and other factors, PERCENT_RANK might be best for this problem.
You really should post the information Centinul requested whenever you have a question.
If you can phrase your question in terms of commonly available tables, then you don't have to post them.
For example, to group employees from the scott.emp table according to their salaries, and get a total of salaries by group, you could create a groups table and join it to the emp table, like this:
CREATE TABLE     groups
(      seq_id     NUMBER
,      label     VARCHAR2 (20)
,      low_pct     NUMBER      -- lowest number IN this group
,      high_pct     NUMBER      -- lowest number NOT IN this group, but in the next highest group
INSERT INTO groups (seq_id, label, low_pct, high_pct) VALUES (1, 'Top 10%',     0,     10);
INSERT INTO groups (seq_id, label, low_pct, high_pct) VALUES (2, '10%-50%',     10,     50);
INSERT INTO groups (seq_id, label, low_pct, high_pct) VALUES (3, 'Bottom 50%',     50,     101);
COMMIT;
WITH     got_pct          AS
     SELECT ename
     ,      sal
     ,      100 * PERCENT_RANK () OVER ( ORDER BY sal DESC
                                              ,          empno
                           )     AS pct
     FROM   scott.emp
SELECT    g.label
,       SUM (p.sal)     AS group_total
FROM       got_pct     p
JOIN       groups     g     ON     p.pct     >= g.low_pct
                      AND     p.pct     <  g.high_pct
GROUP BY  g.label
ORDER BY  MIN (g.seq_id)
;Output:
LABEL                GROUP_TOTAL
Top 10%                     8000
10%-50%                    12875
Bottom 50%                  8150If you need help understanding a query with sub-queries, it helps to run the sub-query by iteself.
In this case:
WITH     got_pct          AS
     SELECT ename
     ,      sal
     ,      100 * PERCENT_RANK () OVER ( ORDER BY sal DESC
                                              ,          empno
                           )     AS pct
     FROM   scott.emp
SELECT     *
FROM     got_pct
;produces this output:
ENAME             SAL        PCT
KING             5000          0
SCOTT            3000 7.69230769
FORD             3000 15.3846154
JONES            2975 23.0769231
BLAKE            2850 30.7692308
CLARK            2450 38.4615385
ALLEN            1600 46.1538462
TURNER           1500 53.8461538
MILLER           1300 61.5384615
WARD             1250 69.2307692
MARTIN           1250 76.9230769
ADAMS            1100 84.6153846
JAMES             950 92.3076923
SMITH             800        100In case of a tie, PERCENT_RANK assigns the same result to all rows with the same input value. The reason I included empno in the analytic ORDER BY clause was to prevent ties. Otherwise, both SCOTT and FORD (with the exact same sal=3000) would be assigned pct=7.69230769, and the "Top 10%" would then include 3 out of 14 rows (over 21% of the rows), which, outside of Lake Woebegone, shouldn't happen.
Edited by: Frank Kulash on Jan 11, 2010 11:02 AM
Explanation added.