How to exclude certain values in a measure.

Question

Hi,
I have fact data something like..
ID
MyValue
MySetValue
1
200
1
2
300
1
3
400
0
4
500
0
5
600
1
Now I want to create a measure MyValue which would be sum(MyValue). This i can do using SUM aggregation.
But I also want a measure which is SUM(MyValue) but excludes rows for which MySetValue is 0.
So first measure should return 2000 and second measure should return 1100.
I know this can be done by adding a new column in view or in DSV which does the filtering. But is there a way of doing the same in calculated member ?
Thanks.
liquidloop[at]live[dot]co[dot]uk

Answer 1

The best way is going to be creating a calculated column in your DSV (or view) and then defining a stored measure based on that calculated column. When business rules are applicable at the grain of the fact table, the best approach is going to be to include
that logic at the fact table. There's ways to create it in the Calculation Tab of the Cube Editor but it's almost never the right way to go. The method that Charles uses above is a devastating bad approach as an aggregation function over a Filter clause though
possible, the SSAS Engine when it sees a statement like those should just raise an error rather than processing the statement. (See point #1 in the following blog:
http://sqlblog.com/blogs/mosha/archive/2008/10/22/optimizing-mdx-aggregation-functions.aspx )The real issue with the approach that Charles uses in your case above is that a Filter statement applies to the current context in aggregate and WILL NOT apply
to each row at the source. Therefore, not only inefficient, it's also incorrect.
To really approach the problem, you would need to create a dimension based on the MySetValue column. In the small example that you have above, that dimension would only have a single attribute hierarchy with  two members but in a real life situation,
that column could and often would be a non-discrete function. Saying that you only have two possible values, and only the single attribute hierarchy with those two members, you could define the new measure as the following:
CREATE
MEMBER CurrentCube.[Measures].[Modified MyValue]
AS
IIf(
              [MySetValue].[MySetValue].CurrentMember
IS [MySetValue].[MySetValue].&[0],
NULL,
              [Measures].[MyValue]
FORMAT_STRING="#,##0";
However, the slice, ([MySetValue].[MySetValue].[All],[Measures].[Modified MyValue] would still be exactly equal to [Measures].[MyValue]. You could fix that using a SOCPE statement.
SCOPE(
       [Measures].[Modified MyValue],
       [MySetValue].[MySetValue].[All]
THIS = (
              [MySetValue].[MySet Value].&[1],
              [Measures].[MyValue]
END
SCOPE;
So as you can see, just do it the right way to begin with. Create a column in your DSV and create a measure from that calculated column. 
Martin
Martin Mason Wordpress Blog