BASICS OF DATA REPRESENTATION The Search For Well- Designed, Useful, and Ethical Graphs.

BASICS OF DATAREPRESENTATION

The Search For Well-Designed, Useful, and Ethical Graphs

DIFFERENT TYPES OF GRAPHS• Useful to know the kinds out there to see if the best type of graph has been

chosen to represent the kind of data being presented.

• Comparison Graphs

• Bar

• Line

• Composition Graphs

• Stacked Area

• Pie Charts

• Distribution Graphs

• Line Histographs

• Relationship Graphs

• Scatter Plots

• Bubble Charts

#1: COMPARISON GRAPHS Show How Variables in a Study Interact

• Useful to show how one or more metrics change over time

• Useful to show correlations and comparisons (such as the rise or Twitter follows after Oscar nominations were announced)

• Useful to compare changes over the same period of time within multiple groups

• Bars are good to show larger changes; Lines are better for shorter changes

COMPARISON GRAPHS (BAR GRAPHS)

COMPARISON GRAPHS (BAR GRAPHS)

COMPARISON GRAPHS (LINE CHARTS)

#2: COMPOSITION GRAPHSShow How Different

Variables Inform a Whole

• Pie Charts excel at showing how percentages of a whole are allocated (multiples can be used for comparison)

• Stacked Area charts excel at showing compositional changes over time along with the Pie Chart’s allocation metric

COMPOSITION GRAPHS (STACKED AREA)

COMPOSITION GRAPHS (PIE)

COMPOSITION GRAPHS (PIE)

COMPOSITION GRAPHS (PIE)

COMPOSITION GRAPHS (PIE)

#3: DISTRIBUTION GRAPHS

Show How and If Correlations Occur

• Best for showing a single variable with A LOT of data points

• One dimension will always be frequency

DISTRIBUTION GRAPHS

DISTRIBUTION GRAPHS

#4: RELATIONSHIP GRAPHS

Show How Two or More Variable Relate to Each

Other

• Scatter Plots show the correlation of two variables with multiple data points

• Bubble Graphs are like scatter plots for three variables but also quickly show amount through the size of the bubble

RELATIONSHIP GRAPHS (SCATTER PLOT)

RELATIONSHIP GRAPHS (BUBBLE GRAPH)

What are Graphs and Infographics Good At?

MAKE SOMETHING MORE ABUNDANTLY CLEAR THAN ONLY WORDS COULD EVER DO

REWORK DATA TO CLEAR UP COMMON CONFUSIONS

Romney/Obama Electoral Map that Could Lead Some to Believe the Country is More Conservative Than Liberal, or Even That Romney Should Have Won.

REWORK DATA TO CLEAR UP COMMON CONFUSIONS

Electoral Map Redrawn to Reflect Population Sizes of Areas.

Now Map More Clearly Shows Obama’s edge (slight as it was).

USING VISUAL RHETORIC TO SHOW BIAS (GOOD OR BAD DEPENDING ON CONTEXT)

Things to Look For, or, Where Graphs Go Wrong.

#1: COULDA BEEN BETTER

#2: ???

#2: LEAVING OUT VALID INFO THAT WOULD HURT YOUR ARGUMENT

#3: STRAIGHT UP LYING

#3: STRAIGHT UP LYING

FBI Supplementary Homicide Reports: 1980-2010 (Known Offenders)

#4: BEING AN IDIOT

#5: ???

#5: NOT SHOWING WHOLE GRAPH (UNLIKE THIS ONE)

#5: NOT SHOWING WHOLE GRAPH

#5: NOT SHOWING WHOLE GRAPH

#5: NOT SHOWING WHOLE GRAPH

#6: FAILING TO PROVIDE A LEGEND (OR A CLEAR ONE)

#6: FAILING TO PROVIDE A LEGEND (OR A CLEAR ONE)

#7: ???

#7: NOT HAVING AXIS GOING IN EXPECTED ORDER AND/ORBEING UNABASHEDLY MANIPULATIVE

#8:NOT PROVIDING REALLY IMPORTANT CONTEXT

#9: ???

#9: UNEQUAL INTERVALS

#10: DESIGN EFFECTS THAT ONLY CONFUSE

#10: DESIGN EFFECTS THAT ONLY CONFUSE

#10: DESIGN EFFECTS THAT ONLY CONFUSE

#10: DESIGN EFFECTS THAT ONLY CONFUSE

OVERVIEW QUESTIONS TO ASK• Has the right type of graph been chosen to represent the point the graph is

trying to make? Or could the graph be re-designed to show the point more clearly?

• Is everything labeled clearly? (x/y axes, colors (if they have meaning), title, source of data included with graph, etc.)

• Is the graph showing enough data points and/or range to mean anything useful?

• Is the graph using equal intervals and thusly not skewing the results?

• If 3D or some other effect has been applied, is it helping or hurting?

• Is the graph leaving out information that could hurt its argument?

• Is the graph using any design/images that show inappropriate bias (this depends on the source)?