Forum Discussion

scrawfor's avatar
scrawfor
Frequent Visitor
4 years ago

User defined scatterplot axes for regression

Long time viewer, first time poster. 

I have a table of facts. There are ~60 columns with ~30,000 rows. The table tracks process parameters. Each row in the table represents a single event. 

I would like to be able to have the report consumer select which column is used on the y-axis and the which is used on the x-axis.

 

There are about 3 columns that act as identifiers (event ID key, facility, date). The remainder are the data that is collected about the event. 

The goal is a user-defined linear regression between any two non-identifier columns that exist in the table.

 

Ive tried using field parameters, unpivoting the table and cross joining to get all possible combinations (this was unworkable), using a copy of the fact table to get the x and y, calculated tables and using a slicer as the way to add columns, etc. 

 

All the usual patterns that work for measures do not work for calculated tables when I tried. 

Any ideas?

2 Replies

  • scrawfor , Final outcome is not clear. But a calculated table can not use slicer values. On a var table inside a measure can be dynamic

    • scrawfor's avatar
      scrawfor
      Frequent Visitor

      Thanks for the reply Amit. I partially solved the problem. Field parameters were the solution.

       

      Sample Data:

       

      Set up two identical Field Parameter tables, one for the x-axis and one for the y-axis:

       

      Created Scatterplot visual and put the "columns" in the x and y-axis fields of the visual.

       

       

       

      Only thing left is to set up the regression. But I cannot refer to the data that underlies the field parameter columns. I need to create measures for the regression calculations {e.g. SUMX(X-Axis, X-Axis[X-Axis]), SUMX(X-Axis, X-Axis[X-Axis] ^ 2), etc.} but the relevant columns of the field parameter tables are strings. Is there any to access the data that the field parameter is referring to? I tried adding these parameter fields to an add-in visual that performs regression calculations and it works so there is a way to access the underlying data set to perform the calculations. I can't use the add-in because I don't have a license and the dataset is limited to 1000 points in the free version. 

       

      Any suggestions?