CSE 5544: Introduction to Data Visualization Raghu Machiraju [email protected] [xkcd] The Visualization Alphabet: Marks and Channels How can I visually represent two numbers, e.g., 4 and 8 How can I visually represent two concepts, e.g., well being and feeling rich … Marks & Channels Marks: represent items or links Channels: change appearance based on attribute Channel = Visual Variable Visualization Families Marks for Items Basic geometric elements 0D 1D 2D 3D mark: Volume, but rarely used Marks for Links Containment Connection Containment - can be nested [Riche & Dwyer, 2010] Channels (aka Visual Variables) Control appearance proportional to or based on attributes Jacques Bertin French cartographer [1918-2010] Semiology of Graphics [1967] Theoretical principles for visual encodings Bertin’s Visual Variables Marks: Points Lines Areas Position Size (Grey)Value Texture Color Orientation Shape Semiology of Graphics [J. Bertin, 67] Image Visual language is a sign system Images perceived as a set of signs Visual Language - System of Signs Sender encodes information in signs Receiver decodes information from signs Jacques Bertin • Image perceived as a set of signs Semiology of Graphics, 1983 • Sender encodes information in signs • Receiver decodes information from signs 13 Semiology of Graphics [J. Bertin, 83] Stolte / Hanrahan “Polaris: A System for Query, Analysis and Visualization of Multi-dimensional Relational Databases”, Chris Stolte and Pat Hanrahan Channel Types: Where / What Based on slide from Mazur Using Marks and Channels Mark: Line Mark: Point Adding Hue Adding Size Channel: Length/Position Channel: Position +1 categorical attr. +1 quantitative attr. 1 quantitative attribute 2 quantitative attr. 1 categorical attribute Redundant Encoding Length, Position and Value Good bar chart? Rule: Use channel proportional to data! Types of Channels Magnitude Channels Identity Channels How much? What? Where? Position Shape Length Color (hue) Saturation … Spatial region … Ordinal & Quantitative Data Categorical Data What Visual Variables? http://www.nytimes.com/interactive/2013/05/25/sunday-review/corporate-taxes.html What Visual Variables ? Characteristics of Channels Selective Is a mark distinct from other marks? Can we make out the difference between two marks? Associative Does it support grouping? Quantitative (Magnitude vs Identity Channels) Can we quantify the difference between two marks? Characteristics of Channels Order (Magnitude vs Identity) Can we see a change in order? Length How many unique marks can we make? Expressiveness + Effectiveness • Expressiveness: – visual encoding should express all of, and only, the information in the dataset attributes – simple one - lie factor • Effectiveness: – importance of the attribute should match the salience of the channel – simple one - data-ink ratio Effectiveness of Mappings - Complex [Kandel et al.: Principles of neural science] Channels: Expressiveness Types and Efectiveness Ranks Magnitude Channels: Ordered Attributes Identity Channels: Categorical Attributes Position on common scale Spatial region Position on unaligned scale Color hue Length (1D size) Motion Tilt/angle Shape Area (2D size) Depth (3D position) Color luminance Color saturation Curvature Volume (3D size) Spatial Location - Position Spatial position 2.05D Position Strongest visual variable Suitable for all data types Problems: Selective: yes Sometimes not available (spatial Associative: yes data) Quantitative: yes Cluttering Order: yes Length: fairly big Example: Scatterplot Position in 3D? [Spotfire] Length & Size Good for 1D, OK for 2D, Bad for 3D Easy to see whether one is bigger Aligned bars use position redundantly For 1D length: Selective: yes Associative: yes Quantitative: yes Order: yes Length: high Example 2D Size: Bubbles Color 36 ????? Color < < Good for qualitative data (identity channel) Selective: yes Associative: yes Limited number of classes/length (~7-10!) Quantitative: no Does not work for quantitative data! Order: no Lots of pitfalls! Be careful! Length: limited My rule: minimize color use for encoding data use for brushing Human Visual System The Eye & Retina Retina Detectors • 1 type of monochrome sensor (rods) – Important at low light • Next level: lots of specialized cells – Detect edges, corners, etc. • Sensitive to contrast – Weber’s law: DL ~ L Retina Detectors 3 types of color sensors - S, M, L (cones) Works for bright light Peak sensitivities located at approx. 430nm, 560nm, and 610nm for "average" observer. Roughly equivalent to blue, green, and red sensors Cone Response HyperPhysics, Georgia State University Color Models RGB Color Space White - Additive system Cyan Yellow - Colors that can be represented by Green computer monitors Blue - Not perceptually uniform Red Black C. Ware, “Visual Thinking for Design” HSL Color Space Hue - what people think of color Saturation - purity, distance from grey Lightness - from dark to light wikipedia.org Not perceptually uniform Thanks to Moritz Wustinger Lab Color Space Perceptually uniform L approximates human perception of lightness a, b approximate R/G and Y/B channels a, b called chroma CIELAB 1976 Perceptual Color Spaces Lightness Colorfulness Hue Unique black and white Courtesy of Maureen Stone Munsell Color Hue, Value, Chroma Value 5 R 5/10 (bright red) N 8 (light gray) Perceptually uniform Chroma Hue Courtesy of Maureen Stone Munsell Atlas Courtesy Gretag-Macbeth Interactive Munsell Tool From www.munsell.com Courtesy of Maureen Stone Another Model - Color Deficiency Color Opponency Red - Green: Difference between R and G Luminance (L): Combination of R and G Yellow - Blue: Difference between L and B C. Ware, “Visual Thinking for Design” Source: M. Stone Model “Color blindness” Flaw in opponent processing Red-green common (deuteranope, protanope) Blue-yellow possible (tritanope -- most common) Luminance channel almost “normal” 8% of all men, 0.5% of all women Effect is 2D color vision model Flatten color space Can be simulated (Brettel et. al.) http://colorfilter.wickline.org http://www.colblindor.com/coblis-color-blindness-simulator/ Color Blindness Simulate color vision deficiencies colorfilter.wickline.org www.vischeck.com Protanope Deuteranope Tritanope No L cones No M cones No S cones Red / green Blue / Yellow Source: M. Stone deficiencies deficiency Color-Blindness Normal Protanope Deuteranope Lightness Source: M. Stone Luminance, Saturation, Hue Luminance How-much channel discriminability: ~2-4 bins contrast important Saturation How-much channel discriminability: ~3 bins Hue What channel discriminability: ~6-12 Value/Luminance/Saturation OK for quantitative data when length & size are used. Not very many shades recognizable Selective: yes Associative: yes Quantitative: somewhat (with problems) Order: yes Length: limited Example: Diverging Value-Scale Color: Bad Example Cliff Mass Color: Good Example Shape Shape ????? < < Great to recognize many classes. No grouping, ordering. Selective: yes Associative: limited Quantitative: no Order: no Length: vast Chernoff Faces Idea: use facial parameters to map quantitative data Does it work? Not really! Critique: https://eagereyes.org/criticism/chernoff-faces Glyphs • Glyphs and icons – Consist of several components • Features should be easy to distinguish and combine • Icons should be separated from each other • Mainly used for multivariate discrete data 68 Glyphs • Color icons [Levkowitz 91] • Subdivision of a basic figure (triangle, square, …) into edges and faces • Mapping of data to faces via color tables • Grouping by emphasizing edges or faces 69 Glyphs Stick-figure icon [Picket & Grinstein 88] 2D figure with 4 limbs Coding of data via Length Thickness Angle with vertical axis 12 attributes Exploits the human capability to recognize patterns/textures Using Stick Figure Icons 71 Glyphs Circular icon plots: Star plots Sun ray plots Follow a "spoked wheel" format Values of variables are represented by distances between the center ("hub") of the icon and its edges Glyphs Star glyphs [S. E. Fienberg: Graphical methods in statistics. The American Statistician, 33:165-178, 1979] – A star is composed of equally spaced radii, stemming from the center – The length of the spike is proportional to the value of the respective attribute – The first spike/attribute is to the right – Subsequent spikes are counter-clockwise – The ends of the rays are connected by a line 73 Texture Mapping to Texture Main parameters for texture Orientation Size Contrast Alternatively [Tamura 78]: Coarseness Roughness Contrast Directionality Line-likeness Regularity [C. Ware, Information Visualization] [C. Ware, Mapping to Texture Goal: Avoid visual "crosstalk“ “Orthogonal” perceptual channels Restricts range of parameters E.g. approximately 30 degrees difference in orientation needed to distinguish textures Main application for textures: nominal data Some applications for direct visualization of orientations Mapping to Texture Generate texture Gabor func. as primitives Parameters: Orientation Visualization] Information Ware, [C. Size Contrast Visualization of a magnetic field Randomly splatter down Gabor functions Blending yields continuous coverage Stochastic texture model Mapping to Texture Other stochastic texture models: LIC (Line integral convolution) for vector field visualization Structural models Procedural description of texture generation E.g. Lindenmayer systems (L- systems) More Channels Other Mappings • More advanced mappings possible • Examples for other visual variables – Motion – Blink coding – Explicit use