Don't use such plots!

Don't use such plots!

In a previous post, I talked about the misuse of box-plots so I decided to go a bit further with this very short article and explain why you should not use two types of plots.

Human eyes and relative areas

The first plot that should be banned is the?pie chart. Indeed human eyes are good at judging linear measures and bad at judging relative areas (slices). Not convinced… I’m pretty sure you write proportions next to each slice. Don’t you? Similarly?bar plots?should rarely be used.

Pie charts should be banned. Use dot charts (or tables).

Good practice is rather to use a table where you display proportions or, if you insist using graphics, a dotchart. If you have several variables where you want to display proportion, a barplot is probably what you need (rather than a table with plenty of rows).

The origin matters

Another mistake is to plot a time series without the origin. Well you can do it but you will lie as any trend will appear much "stronger". Good if you want to show your boss that you advertising campaign had a huge impact but... well I don't have to explain


Closing price for the Apple asset.

PS: An excellent book and much more complete on this topic is Cleveland, W. S. (1985)?The Elements of Graphing Data. Wadsworth: Monterey, CA, USA.

Vincent Tourre

Enseignant Chercheur - Centrale Nantes

1 年

Great! And if you want more, give a look at https://junkcharts.typepad.com/

回复

要查看或添加评论,请登录

Mathieu Ribatet的更多文章

  • CART, Random Forest, Boosting and.... Geostatistics

    CART, Random Forest, Boosting and.... Geostatistics

    I know the title seems a bit odd as geostatistics is far away from 'conventional' machine learning strategies as the…

  • An introduction to Survival Analysis

    An introduction to Survival Analysis

    Lecture series about #statisticalmodelling again (and again). A completely different topic (but very fascinating):…

    3 条评论
  • Advanced Bayesian Statistics and Hierarchical Models

    Advanced Bayesian Statistics and Hierarchical Models

    Some time ago, I posted my lecture materials whose aim was to give an introduction (and a solid theoretical basis) to…

    2 条评论
  • Lecture on Reinforcement learning

    Lecture on Reinforcement learning

    Keep on posting my "series" on #statisticallearning, we are now going to dip a toe in the #reinforcementlearning water!…

  • An introduction to Time series modeling

    An introduction to Time series modeling

    Keep on sharing my mini-courses on various statistical themes, here is my take on Time Series.

  • An introduction to Bayesian statistics

    An introduction to Bayesian statistics

    For those who want to get an introduction to Bayesian statistics, here is my take with a lecture given at Centrale…

    1 条评论
  • New from old (Football model!)

    New from old (Football model!)

    Tomorrow I'll visit Vibracoustic around Nantes to (maybe) start a collaboration. I'm supposed to talk about my research…

  • Brand new lecture on statistical learning

    Brand new lecture on statistical learning

    In a few weeks I will start a new course for the "Option Info" at #CentraleNantes. This course aims at providing some…

社区洞察

其他会员也浏览了