How can I create a flow that uses only select data points (by user markings or table selection) on a scatter plot to calculate and display best fit line without removing the unselected points from the plot?

[Kom Son]

How can I create a flow that uses only select data points (by user markings or table selection) on a scatter plot to calculate and display best fit line without removing the unselected points from the plot?

[Fredrik Rosell]

Hello,

The least complicated solution would likely be if you could have two plots (adding a second plot that serves as a details visualization) - the user marks the data points of interest in plot 1, and the best line fit is shown in plot 2 which uses "Limit data using marking" using the marking from plot 1 to only work on/show those marked values.

[Gaia Paolini]

you could do it in the same plot if you used filtering rather than marking. But the plot would still be split in two sections, one for the filtered data and one for all data. You can do it by Properties > Subsets > then ticking both All data and Current filtering. Example in screenshot

[Fredrik Rosell]

Adding yet another option. Just as Gaia also suggested in the previous comment, I was trying to find a way to use Subsets but thought they had to be based on filtering - not the marking you mentioned. Well, looking at this again, I noticed that you can use Markings as subsets (see screenshot below). Using that, any marked rows would be displayed in a separate Subset.

[Kom Son]

Thanks for your responses. I found that the trend line calculation i.e R2, etc. do not change when using subsets. So I had opted for a details viz or a nested y-axis (both result in updated trend lines) but split plots. I was hoping there'd be a solution for a single dynamic plot based on the underlying script.

[Fredrik Rosell]

Thank you for the update. You mentioned "I found that the trend line calculation i.e R2, etc. do not change when using subsets". I did not see that when testing this (using Logistic Regression Curve Fit) so just curious - what exact Fit Model are you using?

[Kom Son]

straight line fit

[Fredrik Rosell]

Thank you! I just tested again, and was unable to recreate this - it seems to work fine for me using subsets (I tested this using version 12.0 HF.013). If you want to investigate that further, I recommend opening a support case in the Support Portal - https://support.tibco.com - so our support team can get access to more specific information about your environment and analysis to better assist you.