Principles of Data Visualization

Principles of Excellent Visualization

• Informative visualizations provide insight into complex data by exploiting human perceptual pattern matching and recognition capability
• A good visualization should show the data, induce the viewer to think about the substance rather than the methodology or technology of graphic production, and avoid distorting what the data have to say
• A good visualization should also make large data sets coherent, encourage the eye to compare different pieces of data, and reveal the data at several levels of detail

Data and Model Dependency

• A graphical representation can only be as good as the data that is displayed
• An unreasonable model results in an absurd visualization

Principles for Good Quantitative Graphs

• Maintain integrity
• Prevent design variation
• Show data in context
• Maximize data-ink ratio
• Maximize data density

Integrity of Data and Design

• Statistical graphs can be distorted, but graphics are no worse or more vulnerable to distortion than words
• Distortion is introduced in data graphs if the visual representation is inconsistent with numerical values
• Eliminate exaggeration and overdecoration

The Lie Factor

• Violations of the principle of matching graphical extent to represented quantity lead to a Lie Factor
• Lie Factor = size of effect in graphic / size of effect in data
• Lie factors outside [0.95, 1.05] indicate substantial distortion
• Factors of 2 to 5 are not uncommon

Design Variation

• Design variation refers to the change of representation in form or scale over the surface of the graphic
• Leads to ambiguity and deception
• Leads to large lie factors
• Graphic should show data variation, not design variation

Misleading Design

• Visual and statistical tricks geared towards exaggeration of budget growth
• Example: New York Times, February 1, p.IV-7, 1976

Misleading Statistics

• Statistical lapses include ignoring population increase and monetary inflation
• Corrected graph, of the same data, shows steady budget from 1970

Context

• Data out of context provides no comparative scale

Data-Ink

• The larger share of ink on a graphic should present statistical data information
• Data-ink ratio = data-ink used / total ink used
• Maximize the data-ink ratio
• Erase non-data-ink
• Erase redundant data-ink

Clarity of Visual Presentation

• Visualization can quickly become unclear and cluttered
• Guiding principles:
  • Use few and non-obtrusive graphical elements
  • Make good use of the graphical space

Avoid Graphical Clutter

• Decorations and patterns may cause disturbing effects
• Avoid Moiré vibrations

Data Density

• Amount of quantitative data is given by matrix of observed variables
• Data density relates the size of the graphic to the amount of data displayed
• Data density = number of matrix entries / area of graphic

Principles of Excellent Visualization

• Informative visualizations provide insight into complex data by exploiting human perceptual pattern matching and recognition capability
• A good visualization should show the data, induce the viewer to think about the substance rather than the methodology or technology of graphic production, and avoid distorting what the data have to say
• A good visualization should also make large data sets coherent, encourage the eye to compare different pieces of data, and reveal the data at several levels of detail

Data and Model Dependency

• A graphical representation can only be as good as the data that is displayed
• An unreasonable model results in an absurd visualization

Principles for Good Quantitative Graphs

• Maintain integrity
• Prevent design variation
• Show data in context
• Maximize data-ink ratio
• Maximize data density

Integrity of Data and Design

• Statistical graphs can be distorted, but graphics are no worse or more vulnerable to distortion than words
• Distortion is introduced in data graphs if the visual representation is inconsistent with numerical values
• Eliminate exaggeration and overdecoration

The Lie Factor

• Violations of the principle of matching graphical extent to represented quantity lead to a Lie Factor
• Lie Factor = size of effect in graphic / size of effect in data
• Lie factors outside [0.95, 1.05] indicate substantial distortion
• Factors of 2 to 5 are not uncommon

Design Variation

• Design variation refers to the change of representation in form or scale over the surface of the graphic
• Leads to ambiguity and deception
• Leads to large lie factors
• Graphic should show data variation, not design variation

Misleading Design

• Visual and statistical tricks geared towards exaggeration of budget growth
• Example: New York Times, February 1, p.IV-7, 1976

Misleading Statistics

• Statistical lapses include ignoring population increase and monetary inflation
• Corrected graph, of the same data, shows steady budget from 1970

Context

• Data out of context provides no comparative scale

Data-Ink

• The larger share of ink on a graphic should present statistical data information
• Data-ink ratio = data-ink used / total ink used
• Maximize the data-ink ratio
• Erase non-data-ink
• Erase redundant data-ink

Clarity of Visual Presentation

• Visualization can quickly become unclear and cluttered
• Guiding principles:
  • Use few and non-obtrusive graphical elements
  • Make good use of the graphical space

Avoid Graphical Clutter

• Decorations and patterns may cause disturbing effects
• Avoid Moiré vibrations

Data Density

• Amount of quantitative data is given by matrix of observed variables
• Data density relates the size of the graphic to the amount of data displayed
• Data density = number of matrix entries / area of graphic