I have a use case in which each axis (eg: Sulfate, Nitrate in this example) has a different scale. I don't see an easy way to do this. It looks like RadarAxes is based on PolarAxes which fundamentally uses the same scale for x and y dimensions.
I am new to Python and to me it looks like a new RadarAxes projection can be written from scratch (i.e. not based on PolarAxes) which could allow this.
Sorting out a general way to do plotting in a figure between different axis objects is something we should probably do. This relates back to the 'spark line' (maybe have the name wrong) code that @phobson has, but rotated so that the axis spines are radial instead of lined up.
#4699 is adding a lot of improvements to polar plots. Perhaps that could be addressed there or @QuLogic has some ideas having been digging around there a lot lately.
Great stuff all... I guess the question is whether we want to keep this open. The subject line of this issue is still true, however, multiple r scales on a polar plot is hard to display effectively...
I'll ambitiously close this as parallel coordinates plots (PCP) are present in pandas (https://pandas.pydata.org/pandas-docs/version/0.21.0/visualization.html#parallel-coordinates) and I think the original request (later noted as likely worse than a PCP) was to allow radar charts with different scales on each and every coordinates (not just switching r to log), which seems way more complex than #24825 and should probably first be made as a third party library, if anything.
Activity
tacaswell commented on Nov 20, 2015
Sorting out a general way to do plotting in a figure between different axis objects is something we should probably do. This relates back to the 'spark line' (maybe have the name wrong) code that @phobson has, but rotated so that the axis spines are radial instead of lined up.
mdboom commented on Nov 20, 2015
#4699 is adding a lot of improvements to polar plots. Perhaps that could be addressed there or @QuLogic has some ideas having been digging around there a lot lately.
phobson commented on Nov 20, 2015
@tacaswell are you referring to this: http://stackoverflow.com/questions/15167928/implementing-horizon-charts-in-matplotlib?
tacaswell commented on Nov 20, 2015
No it was something with many y axis with lines connecting them what was
used to show rough correlations between high dimensional data points.
On Fri, Nov 20, 2015, 13:53 Paul Hobson notifications@github.com wrote:
phobson commented on Nov 20, 2015
oh oh oh ... this: https://gist.github.com/phobson/9de120cabde660ec734c
By the way -- this turns into a dumpster fire if you try to use log scales.
ankur-gupta commented on Nov 20, 2015
@phobson This plot is much better than a radar chart for my use. Thanks.
mdboom commented on Nov 23, 2015
Great stuff all... I guess the question is whether we want to keep this open. The subject line of this issue is still true, however, multiple r scales on a polar plot is hard to display effectively...
tacaswell commented on Nov 23, 2015
At a minimum @phobson 's demo code should end up in the examples.
phobson commented on Nov 23, 2015
I can take a stab at removing the seaborn and pandas dependencies of the example over the holiday.
anntzer commented on Dec 29, 2022
I'll ambitiously close this as parallel coordinates plots (PCP) are present in pandas (https://pandas.pydata.org/pandas-docs/version/0.21.0/visualization.html#parallel-coordinates) and I think the original request (later noted as likely worse than a PCP) was to allow radar charts with different scales on each and every coordinates (not just switching r to log), which seems way more complex than #24825 and should probably first be made as a third party library, if anything.