custom Discretization patterns

I have a problem with setting a reasonable Discretization pattern on some dimensions.

One of my dimensions has int numbers scattered from 400 to 900. another dimension has float numbers from 0 to 100, however Automatic discretization does not create even ranges for example 0-100,100-200,etc nor 0-10, 10-20, etc. I tried setting DiscretizationBucketCounts to 10,20,100 etc, however, AS does not create rounded bounds for the buckets.

Since the analisys applications are mostly for use by humans, it would make sense to add support for human-readable discretization, like is done in Excel Charts, etc.

Is there a way to setup my own discretization pattern (other than changing the data view to Query with a bunch of Case statements?) ?

From the ASSL documentation in SQL Server 2005 BOL, it looks like an "EqualRanges" discretization method was envisaged, which might have met your needs. Unfortunately, it doesn't seem to have been implemented (as of SP1), at least for DimensionAttribute:

http://msdn2.microsoft.com/en-us/library/ms127037.aspx

>>

SQL Server 2005 Books Online

DiscretizationMethod Element (ASSL)

Defines the method to be used for discretization.

...

Relationship	Element
Parent elements	DimensionAttribute, ScalarMiningStructureColumn

...

EqualRanges

Equivalent to the EQUAL_RANGES discretization method for mining structure columns.

>>

|||

I tried other discretization methods, - Automatic, Clusters, Equal Ranges.

they all seem to group members on odd intervals, like 12.3-13.5, 13.5-14.6, etc.