Forum Discussion

CaptainCrewe's avatar
CaptainCrewe
Frequent Visitor
9 years ago
Solved

Generate a table using SUMMARIZE and GROUPBY

 

Hello

 

As a relative newbie to DAX, I often find myself going round and round getting tantalisingly close to a solution but missing the knowledge to make things work.  To date, head banging has eventually got me a solution, but not today.

 

I have a table of Issuers, the relevant fields being IssuerID, Ticker and Analyst.  The same Ticker can be used across multiple IssuerIDs.

 

I have a table of Assets, with relevant fields of AssetID, IssuerID, Region, Country and NominalAmountUSD.  The join is from IssuerLookup (1) to AssetLookup (Many), on the IssuerID.

 

My task is to show reports by Ticker, Analyst, Region and Country.  An Issuer can have several Assets and hence multiple region/country combinations, but I'm being asked to extract just one to report on.  The business rule is to use the combination that has the highest NominalAmountUSD - that value comes from a Fact table joined to the Asset table.

 

I've learnt how to create a calculated table, and that's a good first step.  The following query lets me group across IssuerIDs that share a Ticker:

 

SUMMARIZECOLUMNS(
    IssuerLookup[IssuerTicker],
    IssuerLookup[Analyst2],
    AssetLookup[Region Of Risk],
    AssetLookup[Country Of Risk],
    "Sum_NV_USD", CALCULATE(SUM(Fact_Assets[NominalAmountUSD]))
)

 

That gives me good groupings, so the next step is to extract the Analyst, Region and Country for a given ticker where Sum_NV_USD is the maximum.  Having made a virtual table from the query above, called IssuerTickerToCountryOfRisk, the following query finds me the maximum SuNV_USD for a given Ticker:

 

GROUPBY(
    IssuerTickerToCountryOfRisk,
    IssuerTickerToCountryOfRisk[IssuerTicker],
    "MaxSum", MAXX(CURRENTGROUP(), IssuerTickerToCountryOfRisk[Sum_NV_USD])
)

 

But I cannot figure out how to combine these to generate a final calculated table that just has the one row per ticker, with the Region and Country coming from the combination with the highest NominalAmountUSD.  I'd like to use CALCULATEDTABLE but the filter argument for that can't use aggregates.  I've experimented with various combinations of ADDCOLUMNS and SUMMARIZE, but without success.

 

I'd like to have a calculated table in the model with these results so as to refer to it from more than one Power BI Visual.  Any guidance on achieving this will be much appreciated.

 

  • Anonymous's avatar
    Anonymous
    9 years ago

    Hi CaptainCrewe,

     

    You can try to use below formula to get the filter result table.

     

    Filtted Table = CALCULATETABLE(Table1,FILTER(ALL(Table1),CONTAINS(Tablel2,Tablel2[IssuerTicker],Table1[IssuerTicker],Tablel2[Max_Sum_NV_USD],Table1[Sum_NV_USD])))

     

     

    Notice: 'table1'(summarizecolumns), 'table2'(group by).

     

    Regards,

    Xiaoxin Sheng

7 Replies

  • Anonymous's avatar
    Anonymous
    Not applicable

    Hi CaptainCrewe,

     

    Can you provide some sample data to test?

     

    Regards,

    XIaoxin Sheng

    • CaptainCrewe's avatar
      CaptainCrewe
      Frequent Visitor

      Thank you for your response.

       

      I'm not sure in what form you need the sample data.  For now, I'll show a couple of extracts from my DAX Studio queries.  The first shows the initial grouping by Ticker, Analyst, Region and Country with a Sum of the NominalAmount to be used as a tie-breaker, obtained from the SUMMARIZECOLUMNS function:

       

      IssuerTickerAnalyst2Region Of RiskCountry Of RiskSum_NV_USD
      AAAJim SmithAfricaMorocco600000
      BBBAmy BloggsNorth AmericaCanada1656090590
      CCCMary DoeSouth AmericaChile894679350
      DDDJoe BrownAsiaChina22909716756
      DDDJoe BrownAsiaHong Kong18898282952
      DDDJoe BrownAsiaMacao1762612500
      DDDJoe BrownAsiaSingapore562370067
      DDDJoe BrownAustraliaAustralia8865500
      DDDJoe BrownEuropeFrance1185432644
      DDDJoe BrownEuropeLuxembourg11250000
      DDDJoe BrownEuropeUnited Kingdom552160000
      EEEMary DoeSouth AmericaChile1293134082
      FFFDavid JonesNorth AmericaCanada3179842513


      I turned this into a calculated table in my model (IssuerTickerToCountryOfRisk) so as to query it more easily and find the MAX(SUM_NV_USD).  

       

      Using the GROUPBY function, I can find the maximum Sum_NV_USD for a given ticker:

       

      IssuerTickerMax_Sum_NV_USD
      AAA600000
      BBB1656090590
      CCC894679350
      DDD22909716756
      EEE1293134082
      FFF3179842513

       

      What I'm aiming for is a calculated table with only one row per ticker.  Where a ticker has more than one combination of Region and Country (ie DDD in this extract), then I need to select the row with the highest Sum_NV_USD.  Easy to do in SQL and I fear I may be bringing too much of that mindset to this problem.  I feel there must be a more elegant way to return the desired rows.

       

      IssuerTickerAnalyst2Region Of RiskCountry Of RiskSum_NV_USD
      AAAJim SmithAfricaMorocco600000
      BBBAmy BloggsNorth AmericaCanada1656090590
      CCCMary DoeSouth AmericaChile894679350
      DDDJoe BrownAsiaChina22909716756
      EEEMary DoeSouth AmericaChile1293134082
      FFFDavid JonesNorth AmericaCanada3179842513

       

      The final objective is to use this summarized table of ticker, analyst, region and country as a replacement lookup table for reports where a single analyst, region and/or country is wanted for a given ticker.  The Sum_NV_USD isn't needed since it can come from a measure calculation over the underlying fact table.

       

      Thanks again for taking an interest

      • Anonymous's avatar
        Anonymous
        Not applicable

        Hi CaptainCrewe,

         

        You can try to use below formula to get the filter result table.

         

        Filtted Table = CALCULATETABLE(Table1,FILTER(ALL(Table1),CONTAINS(Tablel2,Tablel2[IssuerTicker],Table1[IssuerTicker],Tablel2[Max_Sum_NV_USD],Table1[Sum_NV_USD])))

         

         

        Notice: 'table1'(summarizecolumns), 'table2'(group by).

         

        Regards,

        Xiaoxin Sheng