These are chat archives for FreeCodeCamp/DataScience

16th
Dec 2016
evaristoc
@evaristoc
Dec 16 2016 21:11
@becausealice2 I don't really like radar charts, personally. I think suffer from similar caveats as pie charts do. Not saying that I wouldn't use them though: they could be useful to show some differences between mapped attribute values between several different targets when those differences are an evident signature of the targeted phenomenon. But I found them difficult to follow visually and when comparing targets they are not really valuable if there differences are not huge.
Alice Jiang
@becausealice2
Dec 16 2016 21:38
What about for something like this comparing track audio features from Spotify, like "dancability", "acousticness", "liveness" etc.? I've been considering radar or bar charts or is there something else I haven't thought up that would better suit the data?
evaristoc
@evaristoc
Dec 16 2016 21:58

If you don't have many features to compare it could be ok. In fact that is a common example for an application.

Adding: For someone trained in reading them, they are ok and in some situations even welcome. I think it depends on your audience...

Some additional references:
http://blog.scottlogic.com/2011/09/23/a-critique-of-radar-charts.html

The stacked charts/sparklines in this one would be my preferred option...
http://peltiertech.com/spider-chart-alternatives/

I have been using parallel coordinates. Although hard to follow and also requires some training, they offer much more possibilities. It is my impression that they are also more popular than radar charts for interactive projects:
http://duelingdata.blogspot.nl/2013/07/radar-chart-vs-parallel-coordinate-chart.html

My example (just demo with a hack from blocks and using FCC data):
https://bl.ocks.org/evaristoc/b6bb842a4ea3a4fca09d

In general:
If you think they are going to be very confusing, better to report simple numbers instead...
But again, don't discharge the radar ones...

evaristoc
@evaristoc
Dec 16 2016 22:07
Let me know how that went? Going! See you all!
Amelia
@apottr
Dec 16 2016 22:27
@becausealice2 I think if you have one item with several attributes of varying values, I'd say it's a good way to visualize that data
examples:
Alice Jiang
@becausealice2
Dec 16 2016 23:36
Are those PoI screencaps? Lmao
Amelia
@apottr
Dec 16 2016 23:40
well the second one is yeah :P
first one is WestWorld
seems like Jonathan Nolan likes his radar charts :P