4 ms·
The charts have a number of issues. With some changes, they can easily show the significance along with the trend itself. I created a new version of figure 2, w
by nemetroid 7y ago
The charts have a number of issues. With some changes, they can easily show the significance along with the trend itself. I created a new version of figure 2, which fixes some of the issues:
https://svgur.com/i/F35.svg https://svgur.com/i/F35.svg [SVG]
* Bar charts are preferred when presenting a population of something (people or jobs)
* Each Y-axis starts at zero
* Charts that (by definition) add up to 100% are shown as stacked
* The y-axis max is set to 100% where possible (except in one case, where this obscured the trend)
(I couldn't get Google Sheets to make a nice x-axis for the first chart)
- chewbacha 7y agoThank you! The charts really bothered me and this shows that the article is portraying the trends a little more aggressively than they really are. I still think having only two data points (one during the Great Recession and one in 2018) reeks of cherry picking. I’d like the author to have provided more historical data to demonstrate the effect.
- paulsutter 7y agoYour charts could never get published, they’re just not hyperbolic enough > a series of genuine, penetrating shifts have been happening at warp speed through the last decade. These shifts are massively altering the two parties’ economic identities.
- jpster 7y agoThank you! And how did you learn these guidelines? Is there a book or blog post you recommend for data visualization best practices?
- kk58 7y agoNathan yau has a great book
- fact_factory 7y agothis [0] website and the book [1] it references might interest you: [0] https://www.darkhorseanalytics.com/blog/data-looks-better-naked/ https://www.darkhorseanalytics.com/blog/data-looks-better-na... [1] https://www.edwardtufte.com/tufte/books_vdqi https://www.edwardtufte.com/tufte/books_vdqi
- techslave 7y agotufte is the acknowledged reference
- bwb 7y agoVery nice, thanks for creating that!