How is Cluster Discrimination view's Favor propertties calculated?


  • Hi

    I'm writing an article using clustering techniques.

    I have to explain the exact meaning of FAVOR and what is the formula of FAVOR properties?

    I thought that maybe it's calculated based on the following formula:

    Favor for specific field = (count of records for specific state in Cluster1 ) / (count of all records in Cluster1)

    but it's wrong.

    I couldn't find any document about how it's calculated.

    Does anybody know how it's calculated?

    Yashar Zargari

    Monday, October 21, 2013 5:59 AM

All replies

  • The discrimination viewer is always with reference to another category.  The default is "Cluster 1" versus "everything but cluster 1", but you can look at one cluster versus another.  I agree that the calculation is not documented, but I would assume that the total population is the two choices selected (if only two clusters, then the total population of those two clusters, not the whole population).

    The Excel documentation is online with SQL Server 2014, and has pictures:

    Mark Tabladillo PhD (MVP, SAS Expert; MCT, MCITP, MCAD .NET)

    Tuesday, April 22, 2014 10:40 PM