Editing graphics July 10, 2006 Bill Cloud University of North Carolina at Chapel Hill
Jan 02, 2016
Editing graphics
July 10, 2006Bill Cloud
University of North Carolina at Chapel Hill
The agenda
• Game 1: Find the error.
• A quick look at graphics.
• Discussion: How can we reduce errors?
• Game 2: What’s my question?
A good map or chart:A good map or chart:
• Is free of errors.• Is easily understood by the readers.• Makes good use of its allotted
space.• Is clearly connected to the story.• Has text elements that complement,
rather than repeat, accompanying headlines.
A good map …
• Includes, in most cases, both a north pointer and a distance scale.
• Clearly labels the significant elements on the map.
• Labels streets and other landmarks mentioned in the story.
A good chart …
¶ Presents information clearly without wasting space.
¶ Reflects changes in the value of the dollar.
¶ Reflects changes in the population.¶ Appropriately rounds off large numbers.¶ Puts numbers into perspective.
A map should be clear
and uncluttered
(Unlike this one.)
Does this work?Map is cluttered,
but:
•Purpose here is to locate, not to present data.
• Would numbers and a legend (as done here) be more clear?
Accounting for inflation
Stamp prices shown are misleading …
Accounting for inflation
…unless you consider the
changing value of the dollar, as
indicated by the
lengthened lines.
Check inflation at bls.gov
Accounting for inflation
Here’s an interesting chart
showing the growth of the Maine lobster business. The problem is …
Accounting for inflation
Value of lobsters in 2000 dollars
… It doesn’t reflect the change in the value of the dollar
Accounting for inflation
Price Per Pound in 2000 dollars
Actually, the price peaked in 1973!
1950 2002
6.00
$3.00
Accounting for inflation
$60.29 in 2006 dollars.
Linear deception
Linear deception
Is the footnote enough here? Shouldn’t we do something else to show readers that
the ’04 figures are for part of
a year?
Here, change is clearly labeled
• But should we graph such a tentative number?
Linear deception?Does something look
wrong here?
Linear deception?Why are injuries rising and
deaths falling?
1. Deaths are reported deaths; accident figures are estimates, based on a survey of 100 emergency rooms.
2. CPSC estimates actual deaths at figures higher than the reported deaths. Should chart use those numbers?
3. Should we include the 2004 figures at all?
Controlling for population
The chart shows differences in funding amounts among states and some cities, but those differences mainly reflect population differences. Compare it with a graph of the state populations.
Controlling for population
A per-capita comparison presents a very different picture.
Be wary of pie charts
• They do a poor job in ranking proportions.
• The numbers and the slices can be hard to connect.
• The slices get to be too thin for nourishment.
What to do instead?
• Often, simple tables are best.
• Consider the “cake” chart.
•Shares are more clear.
•Rank ordering is simpler. But: What about the number of accidents?
0
20
40
60
80
100
120
140
160
180
1980-81 1990-91 2000-01
Car fatalities, by direction of impact – 1980-2001
Driver deaths in cars 1-3 years old, per million cars
registered
86 (52%)
62 (53%) 41
(46%)
42 (26%)
37 (32%)
32 (37%)
36 (22%)
18(15%)14(17%)
Front Impact
Other
Side Impact
Car safety experts shift focus to side impact
As highway deaths have declined, the share of deaths blamed on side impact crashes has risen.
A lot more information in
the same space.
A Clear ComparisonBut
can the graph show more?
Accounting for the change in population
Worth the space?
(Few data points)
Worth the space?Doesn’t a table work as well?
Worth the space?Where a chart works well
A neat graphic
Lots of data in a limited space
Mistakes of magnitude
True confessions: Errors that made our papers
More true confessions
More true confessions
True confessions, cont’d,
Great catches
Great catches
Great catches
Great catches