|
Bad Charts.
Best's Worst Pie Chart Ever
Backward in Time and Other Errors
From Bowling Alone
Just Plain ChartJunk
More ChartJunk
NYTimes Donuts
Petraeus
testimony
All Clutter:
Time Series Data
Distortion
NYTimes Alphabetical Sorting
Wall Street Journal Data
Distortion
A blogger's bar chart
The Worst Pie
Chart Ever:
Problems with this chart:
Reported by
Joel Best.
For more on numeracy, go to
www.StatLit.org/Numeracy1.htm
Backward in Time and Other Errors:
 |
Problems with this chart:
-
Time is displayed
backwards on the X-axis.
-
3-D effect: does
the scale match the front or the back of the bars?
-
Scaling: the high values
on the community college scale minimize the important
variation in the other two series.
|
Source: Illinois Board of Higher
Education, Illinois Higher Education Annual Report (July 2002), 24
Figure 3.16: Revised Enrollment Chart:
From Bowling
Alone
Problems with this chart:
-
unnecessary 3-D effect.
-
The chart does not
display meaningful data (the most important data are
actually represented by the arrows on the bars, not by the
height of the bar).
-
The chart is not self
explanatory [the letters on the bars reference the textual
explanation].
Source: Robert D. Putnam,
Bowling Alone (New York: Simon and Schuster, 2000), figure 47
Figure 3.5: Revised Chart, with Data from Text:
Just Plain ChartJunk:
Source: Kristin E. Smith, Loretta E. Bass, and Jason M.
Fields. "Child Well-Being
Indicators From the SIPP" Population Division Working Paper No. 24
U.S. Bureau of the Census. Washington, D.C. April 1998
More ChartJunk:
From:
http://himpflcogirdbvb-mbh.freeyellow.com/chart_of_the_week.html
These charts have something to do with futures trading: (click on
thumbnails)
source: :
http://himpflcogirdbvb-mbh.freeyellow.com/chart_of_the_week.html
NYTimes Donuts
Source:
New York Times:
3/14/08: "This
article will appear in this Sunday's [3/16/08] Times
Magazine." |
Problems with this
chart: Charles Blow usually does
a better job with the Times graphics. In this case, he asks the reader to draw comparisons across
what are in effect 18 pie
charts.
Roughly estimating the numbers from
the chart, and just using Obama's share of the primary vote,
this is my revision:
Notes: There is some method to the
sorting here, but it might have been better to sort by the
date of the primary to avoid the implication that the
city\rural divide is increasing.
(and with a smaller font):
 |
Petraeus
testimony
In his Senate testimony defending the progress in
the Iraq war on April 9, 2008, Army General David H. Petraeus
presented several charts, including the one below summarizing the
operational readiness of the Iraqi army and police battalions.
Glenn Kessler, of the Washington Post
comments on several of the charts here:
The stacked bar chart represent the number of
battalions at each stage of readiness, with ORA 1 being the highest
stage.
The problems with the chart are the following:
- Because the most meaningful comparisons
are across time, and not between the National Police and Iraq
Army, it would have been better to construct this chart as a two
panel display, with the three time periods adjacent to each
other, within the police and army panels.
- "The Battalions in the lead" annotation
is undefined. In the first panel it seems to refer to ORA
1 and ORA 2 readiness, but in the second panel, the bracket
stretches down into the ORA 3 level.
- The stacked bars should have been sorted
in reverse order. The March 2008 Army bar is much
taller than the January 2007 Army bar, but mostly because of the
increased number of battalions at the lowest levels of
readiness. Indeed, it looks as if the most crucial
component of the bar -- the dark green at the top --is actually
smaller in 2008 than in 2007.
-
Example of ORA 4 readiness training.
Estimating the numbers from the graphics,
the following chart shows how the "army" side of the chart could
have been constructed. Notice how the relatively modest
increase in the number of battalions at the two highest levels of
readiness.
All Clutter:
 Not
content with the distractions and distortions made possible by the
use of 3-D effects, charters sometimes feel the need to add all
sorts of other Chartjunk to a graph. In the graphics on the
left, Kevin Phillips is trying to make the point that income is
more inequitably distributed in the United States than in other
countries.
Note the extraneous features of this
in this graphic.
- A completely irrelevant map of
the world.
- Two entirely different kinds of
3-D charts displayed at two different perspectives.
- Country names are repeated three
times.
- To display 24 numeric data
points, 28 numbers are used to define the scales.
- The countries are sorted in no
apparent order (not even alphabetically).
- Note the use of the letter " I "
to separate the countries on the bottom chart.
While it might be possible to design a
better graphical display for these data, a table does the job quite nicely:
*Kevin Phillips, The Politics of Rich
and Poor (1991: Harper Perennial), 9.
Time Series Data
Distortion
This is a time series chart originally
printed in a public policy textbook authored by four professors of
political science employed by three public universities.
From these data they conclude:
"There is some evidence that the
cost of higher education may not have escalated so much...
Figure 9-12 reflect the average cost for tuition, room, and
board as a percentage of median family income from 1964 to
1995. While private institutions have increased costs
substantially, public university costs have remained constant.
This indicates that the increased costs associated with higher
education may be quite reasonable when compared to family income
levels." (Cochran, 346-7)
Note the ways in which the authors
have understated the rising costs of public university education.
First, the costs are deflated not by adjusting for the consumer
price index but by median family income -- especially for the years
after 1982, median family income rose much faster than the consumer
price index. Second, graphing both the private and public data on
the same graph enlarges the scale on which the public data is
displayed. It's hard to tell from the graph, but between 1980 and
1995 it appears that public university costs increased from around
11% of family income to near 15% -- in effect, the share of family
income going to public university costs has increased by a third.
The third way of minimizing the cost increases that have occurred
since 1980 is to extend the time series back to 1965.
 A
completely different picture emerges if one were to compare the rate
of increase in public university costs to the rate of increases in
other sectors of the economy. On the left, we see that from 1981 to
1999 -- over the lifetime of today's college student -- public
university costs have risen faster than any other sector of the
economy. Faster even than rising medical care costs. In addressing
the topic of health care inflation, the same authors note that:
"Cost escalation in the medical field has been constant," and
spend four pages of text addressing the reasons for the increases.
(pp. 268-72).
Clarke Cochran et. al. American
Public Policy: An Introduction (1999: St. Martin's Press),
NYTimes Alphabetical
Sorting
Charles Blow is responsible for this chart that
demonstrates two egregious charting errors: the countries are sorted
alphabetically, and the use of the size of bubbles to represent the magnitude of
the data. It's nice that the Times gives so much space to Blow's charts on its
editorial page, but........
charts/Opini3333.pdf
Wall Street Journal Data
Distortion
The side by side charts used in a
Wall
Street Journal editorial, "No Politician Left Behind
Lack of money isn't the problem with education," is a classic example of data
distortion. Note first that the data on the spending is is not adjusted
for inflation or, the growth in the number of pupils. In theory, 500 is
the maximum score on the NAEP scale-scored math tests, but no student ever
reaches this standard. The average score for high school seniors on the
same scale is just over 300.
Including some
more recent data,
and adjusting the reading score scale, we get quite a different picture:
  
A blogger's bar chart
 (click
on chart for full image) |
It takes some time to explain everything
that is wrong with this chart. Fortunately, the esteemed Jorge
Camoes has gone to all the trouble: He explains:
- "No scale: A chart
is about trends and patterns, but you must give the reader at least
some quantitative reference;
- Unnecessary
multiple colors: If you are displaying a single series, it doesn’t
make sense to vary colors by point;
- No sort: sorting
the data establishes a pattern and helps the reader to immediately
see the relative importance of each item;
- Unmanageable
legend: since there are more data points than you can store in your
working memory, the chart forces a pendular movement from the legend
to the chart and back;"
source
|
| 
