How to get sum distinct in the cube. Is it possible.

Question

Here is the scenario.
One report has many countries on it but only one amount.
For a particular day we have the following data in the fact.
TRANSACTION_DAY_NO
Country
Total Amount
19900101
US
34
19900101
IND
35
19900101
IND
36
19900101
AUS
37
19900101
UNKNOWN
38
19900101
UNKNOWN
39
19900101
UNKNOWN
40
19900101
UNKNOWN
41
19900101
UNKNOWN
42
19900101
UNKNOWN
43
19900101
US
43
19900101
IND
42
There are 2 dimensions on the cube.
Date, Country.
I am not sure how to build a cube on this data.
with t as (
select 19900101 transaction_Day_no,     'US' country_no,     34 total_amount from dual union all
select 19900101,    'IND',         35  from dual union all
select 19900101,    'IND',         36  from dual union all
select 19900101,    'AUS',         37  from dual union all
select 19900101,    'UNKNOWN',    38  from dual union all
select 19900101,    'UNKNOWN',    39  from dual union all
select 19900101,    'UNKNOWN',    40  from dual union all
select 19900101,    'UNKNOWN',    41  from dual union all
select 19900101,    'UNKNOWN',    42  from dual union all
select 19900101,    'UNKNOWN',    43  from dual union all
select 19900101,    'US',    43  from dual union all
select 19900101,    'IND',    42  from dual
select transaction_day_no, country_no, sum(distinct total_amount) from t
group by cube(transaction_Day_no, country_no);
I am using AWM. I have tried to build by selecting the following aggregate for the cube
max for the country_no and
sum for the tranaction_Day_no
But i am getting incorrect results.
If i select sum for both country_no and transaction_no then also i get incorrect results.
Please help me solve this issue.
thanks

Answer 1

Thanks for all your reply's.
The problem is that i have duplicates because
One report can have many customers.
One customer can have many countries.
One customer can have many reports.
If i include the report number in the above data and do a sum on both day and report_number and max for everything else then everything is find and i am getting correct results.
But if i take out the report dimension then i am stuffed.
Also the problem is that i can't have one big dimension for the report as the number of reports are in access of 300M
We have tried to solve this issue by having the fullowing.
Dummy Cube.
This has all the combination of all the dimension in the fact table with the report dimension as only one row(-1)
Report Dimension for each Quarter(34M rows each)
Quarter Cube is build.
Then add the values from all the Quarter Cube with the Dummy Cube.
Tried for 2 Quarter and its working fine results are correct as well.
Only problem is that its taking a long time to build the cube because of the report dimension.
I am trying to find a way to remove the report dimension but still use it. As we only use report dimension at level 'ALL'
But if we do aggregation at 'ALL' level the answers are wrong again.
Thanks for looking into this and taking time to reply.
Regards
Alvinder