Customisable Linear Regression Line Scatter Chat
Hi
we have a scatter chart and we are looking to create a line which is Y (Premiums)= X (Claims) so that anything below the line is a loss and anything above the line is profit.
I tried using start from 0 reference line but the premium and claims axis vary i.e. Y goes to 25m and X to 15m so that it doesn't work.
It was suggested a linear regression line may work, but the formula may need to be tweaked, is that possible?
Thanks
Nick
2 replies
-
The reference line is a graphical element. Its not a logical concept that is derived from your data. Using the regression tools, however, is.
At the moment, we do not let you change the underlying regression 'engine' in Discover. You can choose from 4 standard regression techniques (linear, exponential, logarithmic and power) in deriving the regression formula (which is specific to your data set).
The ability to build your own regression engine in Python or R in Discover is on the road map.
If you can resolve it in your data layer, then you can use out-of-the-box regression tools or your own custom code (Python/R) in the yellow Model app instead. Here, you have a blank canvas to do whatever you want.
-
You can create a straight line function in formulate (pic1) and then plot that on a scatter chart. In this example y=2000x
Drag the new measure you created in formulate to the y-values drop zone and have the option to display it as a line. (pic2)
You can even create linear regression based functions here if you want, the standard deviation and correlation functions to construct that are available.
