Sunday 7 February 2010

Why dynamite plots are BAD

Don't use dynamite plots (barcharts displaying mean and std desv). Boxplots and violin plots exist for a reason!!!

Please read the following article about the perils of interpreting dynamite plots
http://biostat.mc.vanderbilt.edu/wiki/pub/Main/TatsukiKoyama/Poster3.pdf

And here a more balance article
http://emdbolker.wikidot.com/blog%3Adynamite

BAD:
A and B seems the same, C and D are the same and A/B are different of C/D

GOOD:
A has a large distribution
B has only TWO POINTS!!
C has an outlier
D Two distributions??







== Example in R ==



[from the Paul Murrel R graphics book]
http://www.stat.auckland.ac.nz/~paul/RGraphics/rgraphics.html




* Barplot:



* Boxplot:
** first the boxplot
** then draw the point with jitter (random shift from the x to avoid overlapping)

No comments: