Pie Charts & Error Bars: Which to Use?

[Jamison Lahman]

Kind of a silly question; can pie charts have error bars?
I have used x-ray dispersion to find the approximate chemical composition of unknown samples. Is it appropriate to have error bars on how much we expect the percentages to deviate statistically throughout the sample?

Or, should I go a different route and use a stacked bar chart? Is it customary to include error bars on those as either?

 

[jedishrfu]

The nearest I've seen to something like this is:

 

[berkeman]

The nearest I've seen to something like this is:

Wow, dat's ugly!

(not your fault jedi)

 

[jedishrfu]

You see these charts a lot in data mining. The inner ring was the subset you decided to analyze and the outer ring would the breakdown over the whole big data set.

and here's one from the FUTURE:

 

[Jamison Lahman]

The nearest I've seen to something like this is:

That's awesome. After looking at this though, I think it'll be best if I just use a regular pie chart and include the uncertainties in the series labels. Thanks

 

[Number Nine]

As an aside, it should be said that there is really no reason to ever use a pie chart, and most people involved in any kind of data visualization despise them. That's not just a personal preference; they're objectively terrible at communicating information. It's almost impossible to directly compare the sizes of two slices unless the differences are very large, and even then you can't read the actual magnitude off of the chart. You're much better off using a bar graph. See e.g. the comparison between pie charts and bar graphs here

http://www.storytellingwithdata.com/blog/2011/07/death-to-pie-charts

and note how you can't actually read the data off of the pie chart (you have to clutter it up with labels to get any information across), but the bar graph immediately communicates the differences between the groups. I actually can't imagine any situation in which a pie chart would be better at communicating information than a bar graph.

That's awesome. After looking at this though, I think it'll be best if I just use a regular pie chart and include the uncertainties in the series labels.

No! What's the point of even having a plot if the viewer just has to get all of the information from text anyway? A pie chart clearly isn't what you want.

 

[Jamison Lahman]

As an aside, it should be said that there is really no reason to ever use a pie chart, and most people involved in any kind of data visualization despise them. That's not just a personal preference; they're objectively terrible at communicating information. It's almost impossible to directly compare the sizes of two slices unless the differences are very large, and even then you can't read the actual magnitude off of the chart. You're much better off using a bar graph. See e.g. the comparison between pie charts and bar graphs here

http://www.storytellingwithdata.com/blog/2011/07/death-to-pie-charts

and note how you can't actually read the data off of the pie chart (you have to clutter it up with labels to get any information across), but the bar graph immediately communicates the differences between the groups. I actually can't imagine any situation in which a pie chart would be better at communicating information than a bar graph.
No! What's the point of even having a plot if the viewer just has to get all of the information from text anyway? A pie chart clearly isn't what you want.

I too have read that people are less equipped to distinguish angular differences. Most of my charts are going to have 2-4 data series so I think a pie chart would be applicable (even if not necessarily the best at conveying the message). Personally, and I am no data visualization expert obviously, I imagine pie charts have one advantage and that is, whatever data is being conveyed, people automatically know it is some type of percentage. Also, I don't mind the labels having numbers anymore than I would if a bar graph had explicit values and a legend :)

I will say, after more thought, I had decided to go with a bar graph but thought it unnecessary to update the post. I plan on comparing numerous sets of data which can more easily be displayed in bars side-by-side. I am still undecided whether I prefer a bar graph to a stacked bar graph so if you know any literature on comparisons of those I wouldn't mind a read :)