10
$\begingroup$

I came across this chart that is both weird and intriguing. It is about some literary works produced in the regions mentioned. The x axis is the time and y axis the percentage.

enter image description here

The preceding para has a hint of how to read but I am still lost.

enter image description here

Reading for Arabia is straight forward, but how about, say, Iraq or Iran? A simple walkthrough will help greatly.

Is this a standard chart or the author's invention?

$\endgroup$
8
  • 1
    $\begingroup$ The bars are stacked and you have to mentally shift them to make sense of the evolution. $\endgroup$ Commented Apr 3 at 7:29
  • 6
    $\begingroup$ aaaaaah! So this is basically a normal stacked bar chart but with vertical lines removed, and areas marked with patterns. Still, the irony in the author's claim is remarkable. I accepted the answer that directly helped me realize the fact, but upvoting other efforts too. $\endgroup$
    – vin
    Commented Apr 3 at 8:13
  • 4
    $\begingroup$ People here will not be surprised but may be saddened at the prospect that calculating percentages is perceived as mathematics that may be distasteful to -- or too challenging for -- the readership. $\endgroup$
    – Nick Cox
    Commented Apr 3 at 12:24
  • 1
    $\begingroup$ Though it is not the primary question you are asking here, I would generally avoid making a stacked bar plot like this for your own uses. There are much better ways to visualize this data (even simply faceting or dodging the bars), and its clear that this method only creates more confusion than is necessary. $\endgroup$ Commented Apr 4 at 1:34
  • 1
    $\begingroup$ @ShawnHemelstrand Agreed. Now, I suspect that the author removed separations between bars to make it easier to indicate the 11 region names directly on them, to try to circumvent the problem posed by using a separate legend (i.e. the number of categories may be already a problem if you have to go back and forth between the bars and the legend - in addition, the "no color" constraint would be another problem for using a separate legend, e.g. look at the trio Arabia/Iran/Egypt). $\endgroup$
    – J-J-J
    Commented Apr 4 at 7:00

4 Answers 4

14
$\begingroup$

Reading for Arabia is straight forward, but how about, say, Iraq or Iran?

The bars are stacked and you have to mentally shift them to make sense of the evolution.

For example the bar for Iraq in 695-743 AD relates to a 75-37 = 38 percent share of the representation and in 1228-1276 AD it relates to a 10-1 = 9 percent share of the representation.

Possibly a different type of graph, like individual bar graphs stacked above each other would have been better. Spain is also directly visible, but the rest is difficult to follow to see if and where they have peaks in their relative influence. (maybe if someone has time to scrape the data and create such graph, this can be demonstrated). An implementation where it may work better is Reebee's chart about the genealogy of popular music.

$\endgroup$
8
  • $\begingroup$ Instead I'd say that in 1228-1776 AD Iraq's share is about 17% - 1% =~ 16%. $\endgroup$
    – rolando2
    Commented Apr 3 at 12:46
  • 1
    $\begingroup$ Now I'm guessing that the other scaleA.H is an Islamic scale, years after Muhammad? $\endgroup$ Commented Apr 3 at 19:22
  • 2
    $\begingroup$ @kjetilbhalvorsen It stands for "Anno Hegirae", i.e. "In the Year of the Hijra" en.wikipedia.org/wiki/Hijri_year . The wikipedia article says: «this era is most commonly denoted as AH (Latin: Anno Hegirae, /ˈænoʊ ˈhɛdʒɪriː/, 'in the year of the Hijra') in parallel with the Christian/Common (AD/CE) and Jewish eras (AM)». You can find other graphs using this kind of double scale (e.g. logarithmichistory.files.wordpress.com/2018/10/… or islamciv.com/wp-content/uploads/2021/07/…), though I have no idea of how common it is. $\endgroup$
    – J-J-J
    Commented Apr 4 at 12:40
  • 1
    $\begingroup$ This kind of double axis scale, where the same value is provided in two different units, is fine with me. The mapping from A.H. to A.D. is not arbitrary. The problematic double axis scales are the ones where two time series are plotted overlaid on one another, with different units, and an arbitrary transformation applied to one of the axes to illustrate the author's point. policyviz.com/2022/10/06/avoiding-the-dual-axis-chart $\endgroup$
    – qdread
    Commented Apr 4 at 16:36
  • 2
    $\begingroup$ The overlaid plotting of multiple time series in a single plot is very common in biotechnology, where the progression of a bioreactor is shown with multiple parameters, biomass, oxygen, CO2 other gasses, temperature, bacterial count, etc. Such cases are not easy to interpret at first glance, but for experienced users of these graphs it is useful. So also for such cases I would be neither for nor against the use. It depends on the public. $\endgroup$ Commented Apr 4 at 19:08
12
$\begingroup$

Yes, this is standard.

The leftmost column says that between 695 and 743 AD, about 38% of biographies were of people from Arabia, another 36% were from Iraq, and the rest from elsewhere.

The rightmost column says that between 1519 and 1568 AD, about 14% of biographies were of people from Arabia, another 29% were from Syria, and the rest from elsewhere.

The order of items on the y-axis doesn’t really matter, but the columns are organized to facilitate comparisons between different periods.

$\endgroup$
9
$\begingroup$

I believe this would be called a stacked bar chart. It shows the proportion that each region makes up in a larger sample of Islamic regions and how this changes across time (broken up by 50-year periods).

It is difficult to interpret at first because it is in grayscale and the authors use a non-intuitive way of describing what is shown.

Iraq's substantial importance under the Abbasid caliphate, for example, is quite clear as is its rapid decline following the Mongol invasions of the 13th century and its virtual obliteration after the time of the Timur around A.D. 1400

You can see this pattern they are describing as Iraq makes up a large proportion on the y-axis from A.D. 695 ~ A.D. 1228, described as "..substantial importance under Abbasid caliphate", and then that proportion becoming smaller after this ("..rapid decline following Mongol invasions of the 13 century..") and continuing to shrink until the end of the x-axis (.."virtual obliteration after the time of the Timur at A.D. 1400").

Here is another example of a stacked bar chart - Stacked Bar Chart.

$\endgroup$
6
  • 2
    $\begingroup$ The graph in the question is composed of vertical bars (not smoothed lines), so I suspect it may be incorrect to use the word "density" here, as there's no kernel density estimation going on (see the definition of "density plot" used by the website you link to r-graph-gallery.com/density-plot.html). Though I'm curious to know what other people think about using the word "density" for describing the graph in the question. $\endgroup$
    – J-J-J
    Commented Apr 3 at 7:32
  • 4
    $\begingroup$ @J-J-J As in other answers I'd just call this a stacked bar chart in which the original data are scaled to percents of total. Other terms include subdivided. I would leave gaps between the bars if I had the data, which appear to be about every 50 years. I wouldn't use the term density here. $\endgroup$
    – Nick Cox
    Commented Apr 3 at 8:09
  • 1
    $\begingroup$ For other methods see e.g. stats.stackexchange.com/questions/56322/… $\endgroup$
    – Nick Cox
    Commented Apr 3 at 8:11
  • 1
    $\begingroup$ @J-J-J I agree, that is my mistake. A stacked bar plot would be the correct term for the graph. I have seen some use the term "density" when proportion is being plotted and "bar" when it is frequency/count, but probably not a correct distinction. $\endgroup$
    – Jack
    Commented Apr 3 at 15:59
  • 1
    $\begingroup$ @Jack thanks for the clarification! As you seem to be a new user on the website, you may not have noticed that you can edit your post if you think it's useful to do so. (Simply click the "edit" link just under your post). $\endgroup$
    – J-J-J
    Commented Apr 3 at 17:19
1
$\begingroup$

Is this a standard chart or the author's invention?

This graph looks like a rather blocky version of the Histomap, created in 1930 by John B Sparks. Sparks was covering a much longer period (four thousand years) so his graph is both aesthetically more pleasing and, in my opinion, easier to follow.

The idea of stacked bar charts goes back much further, for example those by William Playfair (1759-1823), though the information presented is generally simpler.

The first known bar charts (not stacked) were produced by Nicole Oresme in "The Latitude of Forms", in the 14th century.

$\endgroup$
1
  • 1
    $\begingroup$ Stacked or subdivided bar charts are much older, going back in some form or another to at least the 19th century. $\endgroup$
    – Nick Cox
    Commented Apr 4 at 9:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.