1 Reply Latest reply on Oct 16, 2018 7:41 AM by Yuriy Fal

    Correlations again

    Khoma Usire

      Hello good people,

       

      I'm trying to spice my dashboards up with interactive correlation visualizations. Here are a couple of problems I'm hoping you could help me with:

       

      1. I want to correlate two values. That's CORR( [Value1], [Value2] ). But I also want to correlate Value1 in one day to Value2 in the day after. How do I shift Value2 one day further? I can't seem to work out implementation of DATEADD() into CORR().
      2. Is there still no possibility to Correlate values from 2 different sources unless they are joined?
      3. Can I correlate Value1 to Sum of Values2 across some category?

                     As in: I have multiple Values2 for one date, I want to sum them all up and correlate it to Value1 which only exists once per date.

       

       

       

      Thanks ahead,

      Khoma

        • 1. Re: Correlations again
          Yuriy Fal

          Hi Khoma,

           

          As the CORR( expr1, expr2 ) is an Aggregate function,

          its arguments must be evaluated at the Row Level --

          whether they're coming directly from a datasource,

          or being Row-Level Calculations / FIXED LOD expressions.

           

          To calculate a correlation for the Marks on a view --

          at the Visualization Level-Of-Details (VizLOD) --

          one could be using a WINDOW_CORR() Table Calculation.

          Then the arguments for it could be the Aggregates.

           

          Please find the attached (version 10.5 wb)

          as an example of CY / PY monthly correlation.

          Hope it could help.

           

          Yours,

          Yuri