Principles of Information Visualization Tutorial – Part 1 Design Principles

Principles of Information VisualizationTutorial – Part 1Design PrinciplesProf Jessie KennedyInstitute for Informatics & Digital Innovation

Overview! Fundamental principles of graphic design and visualcommunicationØ help you create more effective information visualizations.! Use of salience, colour, consistency and layoutØ communicate large data sets and complex ideas withgreater immediacy and clarity.

Why Visualise?To see what’s in the dataAnscombe’s quartet

Information Visualization! 2 main objectives! Data analysisØ understand the dataØ derive information from themØ involves comprehensivity! CommunicationØ of informationØ involves simplificationJ Bertin, Semiology of Graphics, - Brief Presentation of Graphics, 2004

How do we get from Data to Visualization?! Need to understandØ the properties of the data or informationØ the properties of the imageØ the rules mapping data to imageswww.napier.ac.uk/iidi

How do we get from Data to Visualization?! Need to understandØ the properties of the data or informationØ the properties of the imageØ the rules mapping data to imageswww.napier.ac.uk/iidi

Types of Data! Nominal (labels or types)Ø Sex: Male, Female,,Ø Genotype: AA, AT, AG…! OrdinalØ Days: Mon, Tue, Wed, Thu, Fri, Sat, SunØ Abundance: abundant - common – rare! QuantitativeØ Physical measurements: temperature, expression levelS. S. Stevens, On the theory of scales of measurements, 1946

Data Type Taxonomy! 1D e.g. DNA sequences! Temporal e.g. time series microarray expression! 2D e.g. distribution maps! 3D e.g. Anatomical structures! nD e.g. Fisher’s Iris data set! Trees e.g Linnean taxonomies, phylogenies! Networks e.g. Metabolic pathways! Text and documents e.g. publicationsB. Shneiderman, The eyes have it: A task by data type taxonomy for information visualization, 1996

Types of DataDataTabularOrderedCategorical/NominalMaleFemaleOrdinalAbundantCommonRareQuantitative10 mm15.5 mm23 mmRelationalTreesNetworksSpatialMapsAAATAG…MonTueWed…slide adapted from Munzner 2011, Visualization Principles

How do we get from Data to Visualization?! Need to understandØ the properties of the data or informationØ the properties of the imageØ the rules mapping data to imageswww.napier.ac.uk/iidi

Theory of Graphics! Application of human perceptionØ understand and memorize forms in an imageØ XY dimensions of the plane and variation in Z dimension! Correspondence between data and image! Level of perception required by objective! Mobility or immobility of the imageJ Bertin, Semiology of Graphics, - Brief Presentation of Graphics, 2004

Semiology of Graphics! visual encodingØ points, lines, areaspatterns, trees/networks, gridsØ positional: XY1D, 2D, 3DØ retinal: Zsize, lightness, texture,colour, orientation, shape,Ø temporal:animationwww.napier.ac.uk/iidiSemiology of Graphics. Jacques Bertin, Gauthier-Villars 1967, EHESS 1998

Language of Graphics! Graphics can be thought of as forming a sign system:Ø Each mark (point, line, or area) represents a data element.Ø Choose visual variables to encode relationships between dataelementsdifference, similarity, order, proportiononly position supports all relationships! Huge range of alternatives for data with many attributesØ find images that express and effectively convey the information.www.napier.ac.uk/iidi

Accuracy of Quantitative Perceptual TasksMore accuratepositionlengthangleslopeareavolumeLess accuratedensitycolourCleveland, W.S. & McGill, R. Science 229, 828–833 (1985).

Gestalt Effects! Visual system tries to structure what we see into patterns! Gestalt is the interplay between the parts and the wholeØ “The whole is ‘other’ than the sum of its parts.” – Kurt Koffka! Gestalt Laws/Principles

Principle of Simplicity! Every pattern is seen such that the resultingstructure is as simple as possibleØ Different projections of same cubeØ Perceived as 2 or 3 DØ Depending on the simpler interpretation

Principle of Proximity! Things that are near to each other appear to begrouped together

Principle of Similarity! Similar things appear to be grouped together

Variable Opacity for Clarity! Use of similarity of stroke and opacity to clarify imageØ Layers in the imagewww.napier.ac.uk/iidiM. Krzwinski, behind every great visualization is a design principle, 2012

Principle of Closure! The law of closure posits that we perceptually close up, orcomplete, objects that are not, in fact, completeIllusorywww.napier.ac.uk/iidi

Principle of Connectedness! Things that are physically connected are perceivedas a unit! Stronger than colour, shape, proximity, size

Principle of Good Continuation! Points connected in a straight or smoothly curvingline are seen as belonging togetherØ lines tend to be seen as to follow the smoothest path

Principle of Common Fate! Things that are moving in the same direction appearto be grouped together

Principle of Familiarity! Things are more likely to form groups if the groupsappear familiar or meaningful

Figure-Ground & Smallness! Smaller areas seen as figures against larger background! Surroundednesswww.napier.ac.uk/iidi

Principle of Symmetry! The principle of symmetry is that, the symmetrical areas tendto be seen as figures against the asymmetrical background.www.napier.ac.uk/iidi

3D Effect

Context affects perceptual tasks! Comparing valuesØ LengthØ CurvatureØ AreaØ 2.5D shapeØ Position in 2.5Dwww.napier.ac.uk/iidi

Ambiguous Information: LengthMuller-Lyer

Ambiguous Information: LengthMuller-Lyer

Horizontal-Vertical Illusion

Ambiguous Information: CurvatureTollansky

Ambiguous Information: Area (Context)Ebbinghaus

2.5D ShapeAdapted from Shepard R N, 1990 Mind Sights: Original Visual Illusions, Ambiguities, and other Anomalies (

2.5D ShapeAdapted from Shepard R N, 1990 Mind Sights: Original Visual Illusions, Ambiguities, and other Anomalies (

Ambiguous Information: Position in 2.5D spaceNecker Cube

Preattentive Visual Features! the ability of the low-level human visual system to rapidlyidentify certain basic visual properties! a unique visual property e.g., colour red allows it to "pop out”! aids visual searchingorientationcoloursizewww.napier.ac.uk/iidiAdapted from Perception in visualization C. Healey, : http://www.csc.ncsu.edu/faculty/healey/PP/

Preattentive Visual Features! Some more effective than othersclosurecurvaturelengthwww.napier.ac.uk/iidiAdapted from Perception in visualization C. Healey, : http://www.csc.ncsu.edu/faculty/healey/PP/

Preattentive Visual Featuresflicker direction of movement enclosure/containmentAdapted from Perception in visualization C. Healey, : http://www.csc.ncsu.edu/faculty/healey/PP/

More than 2 Preattentive visual features! A target made up of a combination of non-unique featuresnormally cannot be detected preattentivelyØ spot the red squareØ difficult to detectØ serial search requiredwww.napier.ac.uk/iidiAdapted from Perception in visualization C. Healey, : http://www.csc.ncsu.edu/faculty/healey/PP/

Boundary detectionHorizontal boundaryVertical boundarywww.napier.ac.uk/iidiAdapted from Perception in visualization C. Healey, : http://www.csc.ncsu.edu/faculty/healey/PP/

Region trackingwww.napier.ac.uk/iidi

Use of preattentive features! target detection:Ø users rapidly and accurately detect the presence or absence of a "target"element with a unique visual feature within a field of distractor elements! boundary detection:Ø users rapidly and accurately detect a texture boundary between twogroups of elements, where all of the elements in each group have acommon visual property! region tracking:Ø users track one or more elements with a unique visual feature as theymove in time and space, and! counting and estimation:Ø users count or estimate the number of elements with a unique visualfeature.www.napier.ac.uk/iidi

COLOURCOLOURCOLOURCOLOURCOLOURCO

Colour! “Colour used poorly is worse than no colour at all” -Edward TufteØ “Above all, do no harm”Ø colour can cause the wrong information to stand out andØ make meaningful information difficult to see.

Colour space! A colour space is mathematical model for describingcolour.Ø RGB, HSB, HSL, Lab and LCH! RGB is the most common in computer use,Ø but least useful for designØ our eyes do not decompose colours into RGBconstituents! HSV, describes a colour in terms of its hue,saturation and value (lightness),Ø models colour based on intuitive parametersØ more useful.

Colourimetry! Hue (colour)Ø around the circle! SaturationØ Inside to outsideØ Colour to grey scale! Lightness (value)Ø top to bottomwww.napier.ac.uk/iidiFigs. Courtesy of S Rogers, ONS

Brewer Palettes! Brewer palettes (colorbrewer.org) provide a range of palettes based onHSV model which make life easier for us….Avoid No the perceived use of order hue toencode of importance quantitative variablesQuantitative encodinge.g. heat mapsTwo-sided quantitativeencodingswww.napier.ac.uk/iidiFig. Courtesy of M. Krzwinski,

ExamplesPoor useof colourBrewercolourswww.napier.ac.uk/iidiM. Krzwinski, behind every great visualization is a design principle, 2012

Conversion to Grey scale! Ensure chosen colour set works well in grey scaleØ Sequential palette works well herewww.napier.ac.uk/iidiFig. Courtesy of M Krzywinski

Trouble with perceptual colour….www.napier.ac.uk/iidiFigs. Courtesy of S Rogers, ONS

Context Affects Perceived Colourwww.napier.ac.uk/iidiFigs. Courtesy of S Rogers, ONS

Colour & Accessibility….Accessibility (W3C):10-20% of population arered/green colour blind.(74? 21? No number atall?)….www.napier.ac.uk/iidiFigs. Courtesy of S Rogers, ONS

Colour Blindness8% males ofUSA descentRed-greenRed-greenBlue-yellowwww.napier.ac.uk/iidiFig. Courtesy of M Krzywinski

BioVis Example:Immunofluorescenceimagesred-green image ofP2Y1 receptor andmigrating granuleneurons,effectively remappedtomagenta-green usingthe channel mixingmethod.www.napier.ac.uk/iidiFig. Courtesy of M Krzywinski

! Blue-YellowØ might bebetter than! Green-MagentaØ talk aboutsame colourswww.napier.ac.uk/iidiGabriel Landini & D Giles Perryer, Image recolouring for colour blind observers

From Data to Visualization…! The properties of the data or information! The properties of the image! The rules mapping data to imageswww.napier.ac.uk/iidi

Encoding SchemespositionslopedensitylengthareasaturationangleshapehueconnectioncontainmenttextureAdapted from Mackinlay J (1986) Automating the design of graphical presentations of relational information.

Mapping data types to encodingMackinlay J (1986) Automating the design of graphical presentations of relational information.

Don’t forget Salience…! Physical properties that set an object apart from itssurroundingsØ Distinct features have high salience! Encodings have differences in discrimination and accuracy! Context affects salience! Choose salient encodings for primary navigationØ Colour is good for categories - salience decreases with more hues.! Focus attention by increasing salience of interesting patterns! Unexpected or bad things can happen when unimportantelements in a figure are salient.Ø The reader will use salience to suggest what is important.www.napier.ac.uk/iidi

Example Encodings in BioVis…DNA sequence – 1D, Nominal data, colourMicroarray gene expression –2D, Ordinal data, colour, positionPhylogeny –Tree, Nominal data, position, colourMicroarray time-series –temporal data, quantitative data,Position, height, colour

ExamplesSequence alignments – matrix, colour,position, lengthUse of Symmetry, hue, saturation, lengthMarshall et al http://bioinf.hutton.ac.uk/tablet/screenshots.shtmlBorkin et al, 2011 Evaluation of artery visualizations for heartdisease diagnosis

Examples! CircosØ human genomeØ location of genesimplicated in diseaseØ regions ofself-similaritystructural variationwithin populationsØ Uses:links, heat maps, tiles,histogramsUse of colour, goodcontinuity, length,transparency, ..www.napier.ac.uk/iidiKrzywinski http://circos.ca/intro/genomic_data/

Which Encoding?! Challenge:Ø Pick the best encoding from the large number of possibilities.Ø Wrong visual encoding can mislead or confuse user! Visual Representation should be expressive! Principle of Consistency:Ø The properties of the representation should match the properties ofthe data.! Principle of Importance Ordering:Ø Encode the most important information in the most “effective” way.www.napier.ac.uk/iidi

Expressiveness! Visual Representation encodes all the factsØ An nD (or 1:N) data set e.g. Iris data setØ cannot simply be expressed in a single horizontal dot plot becausemultiple cases are mapped to the same positionwww.napier.ac.uk/iidiFig. Courtesy of M Krzywinski

Expressiveness! Encodes only the factsØ Wrong use of a bar chart implies something better about Swedishcars than USA ones…("011*.2"Sweden '"3)"4516."0+27"'!!!"France &"839"%$!7"):5;

Consistency! Visual variation in a figure should always reflect and enhanceany underlying variation in the data.! Avoid using more than one encoding to communicate thesame information.! Do not use visually similar encodings for independentvariableswww.napier.ac.uk/iidi

Consistency! red - processed genes, but salience attenuatedØ other genes encoded with competing glyph - red star.www.napier.ac.uk/iidiM. Krzwinski, behind every great visualization is a design principle, 2012

Consistency! Uniform size and alignment of exons and introns reducescomplexity and aids interpreting their complex arrangement.Sharov AA, Dudekula DB, Ko MS (2005) Genome-widewww.napier.ac.uk/iidiassembly and analysis of alternative transcripts in mouse.Genome Res 15: 748-754.M. Krzwinski, behind every great visualization is a designprinciple, 2012

Consistency! Order items in alegend according toorder ofappearance in theplotwww.napier.ac.uk/iidiM. Krzwinski, behind every great visualization is a design principle, 2012

Consistency - Navigational aids! Use consistent axes when comparing chartswww.napier.ac.uk/iidiRaina SZ, et al. (2005) Evolution of base-substitution gradientsin primate mitochondrial genomes. Genome Res 15: 665-673.M. Krzwinski, behind every great visualization is a designprinciple, 2012

Increase data:ink ratio! Navigational aidsØ should not competewith the data forsalience.! AvoidØ heavy axes,Ø error bars andØ glyphswww.napier.ac.uk/iidiHeer J, Bostock M (2010) Crowdsourcing graphical perception: using mechanical turk to assess visualization design.Proceedings of the 28th international conference on Human factors in computing systems. Atlanta, Georgia, USA: ACM. pp. 203-212.

Increase data:ink ratio! Avoidunnecessarycontainmentwww.napier.ac.uk/iidiFig. Courtesy of M Krzywinski

Increase data:ink ratio! Avoid “Chart junk”Sharov AA, et al (2006) Genome Res 16: 505-509.Peterson J, et al. (2009) Genome Res 19: 2317-2323.www.napier.ac.uk/iidiThomson NR, et al. (2005) Genome Res 15: 629-640.DB, Ko MS (2005) Genome Res 15: 748-754.M. Krzwinski, behind every great visualization is a designprinciple, 2012

Keep things simple - Avoid 3D! 3D scatter plots are better as a series of 2D projections.www.napier.ac.uk/iidiSon CG, et al. (2005) Database of mRNA gene expression profiles ofmultiple human organs. Genome Res 15: 443-450.M. Krzwinski, behind every great visualization is adesign principle, 2012

Interactivity…Beyond Basic Design: Interaction! The potential to overcome well known problems withstatic imagery....

Change Blindness….SPOT THE DIFFERENCEwww.napier.ac.uk/iidiFig. courtesy of S Rogers, ONS, UK

Interaction! Supports the user in exploring data! Shneiderman’s Information Seeking Mantra:Ø Overview first, zoom and filter, then details on demandwww.napier.ac.uk/iidi

Interaction: operations on the data! sorting! filtering! browsing / exploring! comparison! characterizing trends and distributions! finding anomalies and outliers! finding correlation! following pathwww.napier.ac.uk/iidi

Interaction: Techniques to support operations! Re-orderable matrices - sorting! Brushing - browsing! Linked views – comparison, correlation, different perspectivesØ Linking! Overview and detail -Ø Excentric labelling! Zooming – dealing with complexity/amount of data! Focus & context - dealing with complexity/amount of dataØ Fisheye….Ø Hyperbolic! Animated transitions - keeping context! Dynamic queries - exploringwww.napier.ac.uk/iidi

Sortable tables, matriceswww.napier.ac.uk/iidi

Linked Views, Linking, Brushing, ExcentricLabels! Time-seriesMicroarrayDatawww.napier.ac.uk/iidi

Focus and Context – Fisheye treewww.napier.ac.uk/iidi

Animated transitions, brushingwww.napier.ac.uk/iidi

Drill Down, Select, Zooming, Focus & Contextwww.napier.ac.uk/iidi

Brushing, Filtering, Principle of Continuitywww.napier.ac.uk/iidi

Linked Views, Filteringwww.napier.ac.uk/iidi

Multiple Trees Comparison! Brushing! Focus &contextwww.napier.ac.uk/iidi

Zooming, filterwww.napier.ac.uk/iidi

Smoothly Animated Transitionswww.napier.ac.uk/iidi

Principles of InformaitonVisualisation Tutorial :Part 2 - Design ProcessCydney Nielsen