Skip to content
 

Suspicious histogram bars

Aleks sent me this (I’m not sure from where):

4 Comments

  1. Truman says:

    It would be much better if the error bar sizes were reversed

  2. Jeff Eldridge says:

    Andrew, is this really a histogram? In any case, a Google image search turns up this post as the possible source:

    http://velica.deviantart.com/art/Error-bars-101948712

    The artist recently revisited the original drawing:

    http://velica.deviantart.com/art/To-err-is-human-272559879?q=gallery%3Avelica&qo=0

    and this revamp is available as a T-shirt:

    http://www.redbubble.com/people/velica/works/8167432-to-err-is-human

  3. Nick Cox says:

    Many biostatisticians and some others call this a dynamite plot. I wouldn’t call it a histogram; the implication is that we have estimates and associated error bars.

    See e.g. http://biostat.mc.vanderbilt.edu/wiki/Main/DynamitePlots

    @Truman: No. Looks the right way round to me: the bar with the biggest error bar is least trustworthy.

  4. [...] had a good laugh when I saw Andrew’s reference to this [...]