Options

Show LOESS in the scatter plot

yzanyzan Member Posts: 66 Unicorn
edited December 2018 in Product Feedback - Resolved

One of the important purposes of visualizations is estimation of the effect of an attribute to the label. One of the basic visualizations for that purpose is a scatter plot. But sometimes, the effect is difficult to inspect visually. For that purpose other tools like SPSS, SAS or Orange provide ability to add a trendline. Even Excel has this feature.

 

My proposal:

Add a checkbox into scatter chart that says something like: "Show LOESS".

 

Reasoning:

  1. LOESS is a reasonable non-parametric method, which can deal with non-linearities.
  2. It is already implemented in apache.commons.math3 library, which is already used by RapidMiner.

Implementation details:

  1. Show confidence intervals, not just the fitted curve, if possible.
  2. If "Color Column" is defined and is discrete, show a fit for each colour. Something like Mondrian does. Including the transparency of the confidence intervals.
  3. Possibly respect sample weights, if specified.
3
3 votes

Fixed and Released · Last Updated

9.2

Comments

  • Options
    sgenzersgenzer Administrator, Moderator, Employee, RapidMiner Certified Analyst, Community Manager, Member, University Professor, PM Moderator Posts: 2,959 Community Manager
Sign In or Register to comment.