effective data presentation: a workshop a.5 ‘rules’ for graphical excellence b.9 common types of...

40
Effective data presentation: a workshop ‘rules’ for graphical excellence Common types of graphs/ figures (and errors and biases in each) ps for improving data display Robert Rhew Department of Geography Dept. Environmental Science, Policy & Management [email protected]

Upload: morgan-walters

Post on 29-Dec-2015

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Effective data presentation: a workshop

A. 5 ‘rules’ for graphical excellence

B. 9 Common types of graphs/ figures (and errors and biases in each)

C. Tips for improving data display

Robert RhewDepartment of GeographyDept. Environmental Science, Policy & [email protected]

Page 2: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Principles of graphical excellence

Graphical excellence…

…is the well-designed presentation of interesting data- a matter of substance, of statistics and of design

…consists of complex ideas communicated with clarity, precision and efficiency

…is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space.

… is nearly always multivariate

… requires telling the truth about the data.

(E. Tufte, The Visual Display of Quantitative Information, 2nd ed)

Page 3: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

B. Common types of graphs/figures

1. pictograph

2. pie charts

3. bar charts

7. scatterplot (generally, relational graphics)

8. narrative graphics of space and time

4. line chart / time series

5. maps

6. histogram/ph

Page 4: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

A. Five ‘rules’ for graphical excellence

1. Show the data

2. Tell the truth (about the data)

3. Maximize data-ink

4. Minimize chart-junk

5. Have a message

Page 5: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Ways to violate the rules

1. Don’t Show the data

2. Don’t Tell the truth

3. Don’t Maximize data-ink

4. Don’t Minimize chart-junk

5. Don’t Have a message

Examples?

Page 6: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

1. Pictographs (pictograms)…use picture symbols to convey the meaning of statistical information.

% of people who will never take you seriously again“How Your Health Care Dollar

Is Spent” (Blue Cross& Blue Shield of Michigan)

Using pictures, though, is not the same as a pictograph

Page 7: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Pictographs are prone to bias and “chart junk”

1960: $1.00

1971: $1.60

1980: $3.10

1990: $3.80

Minimum Wage

www.health.uottowa.ca/biomech/courses/hss2381graphs.pdf

Page 8: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

2. Pie charts

… summarize categorical data, % distributions, share of total

1. use clear and descriptive titles, labels, data points

2. refrain from using more than one pie chart per figure

3. Include properly labeled slices (better than legends), total amount, no overcrowding.

4. Tip: Six slices maximum5. Tip: Order segments by size

(largest to smallest) and in a clockwise direction.

Pies I’ve eaten recentlyPumpkin 14Apple 11Mixed berry 5Cherry 3Blueberry 1

Page 9: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

3-D pies can bias a slice depending on position: put in back to reduce its size; slice in front makes it larger; separating makes emphasis

Which pie graph presents the least biased view of “other province student enrollments”?

source: U Ottawa

“A table is nearly always better than a dumb pie chart…pie charts should never be used.” (Tufte, p. 178)

Page 10: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Counterargument: pie charts have their place

Chlorine equivalents in the atmosphere

WMO, 2008

Page 11: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

3. Bar charts

Used to show: comparison between items, differences, ranks.

•Include clear and descriptive titles for labels and axes

•limit clutter: limit number of bars

• Use color effectively, not gratuitously.

• Careful: y-axes, temptation to use 3-D graphics, inappropriate use (2 variables), improper binning.

Summary of the principal components of the radiative forcing of climate change.

Page 12: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

IPCC Climate Change 2007

What improvements were made?

Page 13: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Potential problem: No Relative Basis

FR = Freshmen, SO = Sophomore, JR = Junior, SR = Senior

What is the message of this plot? What about this plot? (same data set)

Example: “Younger students do better in my class: most of the A’s go to freshmen and sophomores.”

Page 14: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

4. Time series / line charts

Used to show fluctuations or variations over time

• Include beginning and end dates

• clear and descriptive titles for labels and axes

• data points (within reason)

• BE CAREFUL: lines suggest

continuity and linkage

Page 15: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

What’s wrong with this graph?

(slide courtesy of Sylvia Bunge, UC Berkeley)

Page 16: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Here, there’s progression from pre-training to post-training. There is no progression from French to English speakers

Lines indicate a link, a progression

(slide courtesy of Sylvia Bunge, UC Berkeley)

Page 17: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Source: IPCC, Climate Change 2007, Fig TS.14

5. Maps

Used to show regional variation, patterns (geography)• use clear and descriptive labels• include data points (within reason)• limit # of colors to no more than 4. (Use colors to help viewer, lightest to

darkest, rather than a hodgepodge of greens, reds, blues and yellow).

Page 18: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Even maps are susceptible to biased presentations. Why?

http://www-personal.umich.edu/~mejn/election/

Page 19: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Cartogram: map in which sizes of states rescaled according to their population.

http://www-personal.umich.edu/~mejn/election/

Page 20: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

scaling size of states to be proportional to # electoral college votes

http://www-personal.umich.edu/~mejn/election/

Page 21: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Same thing with county-level election results

http://www-personal.umich.edu/~mejn/election/

Page 22: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Cartogram of county-level election results

http://www-personal.umich.edu/~mejn/election/

Page 23: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Vanderbei: red, blue, shades of purple

http://www-personal.umich.edu/~mejn/election/

Page 24: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Cartogram

http://www-personal.umich.edu/~mejn/election/

Page 25: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

6. Histograms…summarize discrete or continuous data that are

measured on an interval scale… are generally used when dealing with large data sets (>100

observations). Histograms can also help detect outliers / gaps.

A vertical bar graph and a histogram differ in these ways: * Histogram: frequency is measured by the area of the column. * Vertical bar graph: frequency is measured by the height of the bar.

“In Their Prime, and Dying of Cancer” Science 317, 5842, p. 1160-62 (Aug 31, 2007)

From 1975 to 1999, the chance of surviving cancer for 5 years slowly improved in older adults and children but not for those in between.

Page 26: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

7. Scatterplots “the greatest of all graphical designs” (Tufte, p. 47)

… present measurements of 2 or more related variables. Useful when variable on y-axis thought to be dependent on value of x-

axis (usually an independent variable).

abacus.bates.edu

Page 27: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Consider hypothesized directionality of causation

Independent variable goes on the x-axis

(slide courtesy of Sylvia Bunge, UC Berkeley)

Page 28: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

projections (time) of regional temperature increases (space) given estimated future greenhouse forcings

Source: IPCC, Climate Change 2007, Fig TS.28

SRES scenario

B1

A1B

A2

8. Narrative graphics of space and time

Page 29: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Phylogeny and body size change within paravian theropods. A temporally calibrated cladogram

A Basal Dromaeosaurid and Size Evolution Preceding Avian FlightTurner et al., Science 317, p. 1380 (2007)

Page 30: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

SEM of tree rings from tiny piece of ancient Siberian pine. Narrow, deformed rings at center correspond to 536 to 537 A.D. and graphically record a catastrophic summer cooling that froze the tree's sap, and that maybe linked to a massive eruption of a young volcano, the precursor to Krakatoa.

Dee Breger, Science, 301, 1472-1473 (2003)

Mongolian Frost Rings (1st place)

9. A photo can speak 1000 data points

Page 31: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

Principles of analytical design

1. Comparisons: show comparisons, contrasts, differences

2. Causality, mechanism, structure, explanation3. Multivariate analysis: show multivariate data; that is,

show more than 1 or 2 variables.4. Integration of evidence: completely integrate words,

numbers, images, diagrams5. Documentation: thoroughly describe the evidence.

Provide a detailed title, indicate the authors and sponsors, document the data sources, show complete measurement scales, point out relevant issues.

6. Content counts most of all: analytical presentations ultimately stand or fall depending on the quality, relevance, and integrity of their content.

Tufte, Beautiful Evidence, 2006

Page 32: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

C. Tips for improving data display

1. Include error bars2. A word on means3. Use color carefully and sparingly4. Maximize Data-Ink5. Minimize Chart Junk6. Avoid moiré effects7. Reduce spaces and grid lines8. Use shading effectively9. The shape of the graphic10.Other thoughts

Page 33: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

1. A word on error bars

IPCC, Figure TS.22

• Display errors - in papers, talks, and even when exploring the data

• But avoid obscuring data with your error bars

Page 34: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

2. Means don’t tell the whole story

Smurfs Turfs

(slide courtesy of Sylvia Bunge, UC Berkeley)

Page 35: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

3. Use color carefully & sparingly

http://en.wikipedia.org/wiki/Color_blindnesshttp://colorvisiontesting.com/what%20colorblind%20people%20see.htm

In the United States, about 7 percent of the male population - or 21 million men - and 0.4 percent of the female population either cannot distinguish red from green, or see red and green differently (Howard Hughes Medical Institute, 2006).

Normal Protanope Deutanope

Page 36: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

4. Maximize data ink 5. Minimize chart junk

Tufte, p. 94data-ink: the non-erasable core of a graphic, the non-redundant ink arranged in response to variation in the numbers presented

data-ink ratio =data-ink

total ink used to print the graphic

Page 37: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

6. avoid moiré effects (unless you are an advertiser)

“design interacts with physiological tremor of the eye to produce distracting appearance of vibration and movement.”

distracting and chartjunk

Tufte, p.111

Page 38: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

7. Reduce spaces and grid lines 8. Use gray shading effectively

sort of like this, but better

Page 39: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

9. The shape of the graphic1. if the nature of the data suggest a shape of the

graphic, use that shape

2. Otherwise choose a horizontal graphic about 50% wider than tall

longer horizontal helps

effect

elaborate causal variable

text kind of far

apartcause

Tufte, p. 187

slopes best detected around 45°

Page 40: Effective data presentation: a workshop A.5 ‘rules’ for graphical excellence B.9 Common types of graphs/ figures (and errors and biases in each) C. Tips

References and sources of materialTEXTSTufte, E. The Visual Display of Quantitative Information, 2nd edition. 2001. Tufte, E. Beautiful Evidence, 2006

ONLINEStatistics Canada (http://www.statcan.ca/english/edu/power/ch9/using/using.htm)

homepages.ius.edu/SRAUSC01/Chap2.html

www.health.uottowa.ca/biomech/courses/hss2381graphs.pdf

Kaiser Family Foundation (http://www.kaiseredu.org/tutorials/ effective_graphics/ alantutorial_updatedbuttons.html)

Bates College Online Resources for Biology: http://abacus.bates.edu/~ganderso/biology/resources/writing/HTWtablefigs.html#scatterplot