1
$\begingroup$

I am creating a customizable analysis in which a requested feature is a radar chart, which I have been powering using Plotly in Javascript. My issue is that I must now be able to output a Radar chart (or similar visual) if only two variables are relevant. Radar Charts really aren't suitable for less than three variables, so I'm curious what people have used as alternatives.

I've considered adding an empty dummy variable to give the dimensionality, and then crop my chart to only show the variables of interest (below).

Solution for Radar Chart visualization with only two variables

I'm curious if people who have dealt with this have better implementations or suggestions, while keeping in mind they should be similar to radar charts. For example, please don't answer with a histogram or other visualization that is much different from the radar chart concept.

Edits:

I standardize the variables before plotting. My primary intention with this visualization is for someone to quickly tell how the variables are performing with reference to the "safe range" (the orange shading)

There will only be one min, one max, and one average per variable.
The raw data is from time series, each variable is a time series of said variable.

$\endgroup$
5
  • 2
    $\begingroup$ sometimes you just need to set your foot down to the powers that be and say "this is not the appropriate visualization here, you should use ____" This radar chart is absolutely garbage for comparing 2 values. I cannot easily tell from a glance even which variable is larger. $\endgroup$
    – bdeonovic
    Jan 13, 2022 at 15:30
  • 1
    $\begingroup$ I appreciate the sentiment and your input @bdeonovic, as to your last point I could stop hiding the units but I standardize the variables before plotting. My primary intention with this viz is for someone to quickly tell how the variables are performing with reference to the "safe range" (the orange shading), not necessarily for A ><= B $\endgroup$
    – jros
    Jan 13, 2022 at 15:34
  • $\begingroup$ I think thats important information to include in the original post. $\endgroup$
    – bdeonovic
    Jan 13, 2022 at 16:01
  • $\begingroup$ are these variables jointly measured? ie i see you have multiple observations (hence avg, min, max) is it the case that for 1 observation you get variable A and B? $\endgroup$
    – bdeonovic
    Jan 13, 2022 at 16:03
  • $\begingroup$ @bdeonovic you're right, I've added these specifications in the Edits section of my post $\endgroup$
    – jros
    Jan 13, 2022 at 16:25

1 Answer 1

0
$\begingroup$

How about rectangles like this:

enter image description here

You could also calculate some summary statistic like the overlap of the intersection of these 2 rectangles. You could add marginal boxplots to show more information about the distributions of each variable. Ideally you would have data from the joint distribution of A and B and then you could just do contour plots

$\endgroup$
7
  • 1
    $\begingroup$ Could you explain the way in which such a graphic conveys the information described in the question? The visual metaphor isn't apparent to me. $\endgroup$
    – whuber
    Jan 13, 2022 at 17:07
  • $\begingroup$ the graphic shows the boundaries of the variables A and B along with the boundaries of the "safe" zone for both of these variables. You can quickly at a glance see if the red rectangle is within the blue one. $\endgroup$
    – bdeonovic
    Jan 13, 2022 at 17:24
  • $\begingroup$ Yes, the rectangle inclusion idea is clear, but I still have no clue what is being shown in this graphic. What are the data and how are they represented? What would be the radar plot equivalent of this graphic? $\endgroup$
    – whuber
    Jan 13, 2022 at 18:32
  • $\begingroup$ The radar chart shows 3 summary values for each variable (min, avg, max) and then also min and max for "safe" region. The rectangle plot shows the exact same thing I just excluded a point for average for variables A and B. $\endgroup$
    – bdeonovic
    Jan 13, 2022 at 19:21
  • $\begingroup$ Thanks. Curiously, because the "safe range" limits in the example radar charts are linear, back in Cartesian coordinates--as here--that range should be strongly curved. Also, no minima appear in any of the radar chart legends. $\endgroup$
    – whuber
    Jan 13, 2022 at 19:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.