INF 252 - Visualization - Information visualization

What is scagnostics?

Scagnostics (scatterplot diagnostics) is a concept used to identify key characteristics or features in a scatterplot, such as clusters, outliers, gaps, and trends. It uses a set of nine measures to provide a summary of the scatterplot, helping to understand the underlying distribution or relationship between variables.

What is the purpose of a scatter plot in data visualization?

Scatter plots are used to visualize trends, communicate outliers, distributions, clusters, and correlations. They are particularly effective for visualizing two quantitative attributes.

What is the limitation of scatter plots?

Scatter plots become less effective when they become too dense. Their scalability is often limited to a few hundred or maximum thousands of items.

Set'o'Gram, Parallel Sets, Radial Sets

Set'o'gram: Each set is represented by a bar broken up into blocks of varying width Parallel Sets: Axis layout of parallel coordinates where Boxes are categories Radial Sets: For large number of items. • Sets: Radially arranged regions • Overlaps: links between regions (need to be interactive)

Sugiyama layout steps and pros/cons

Step 1 - Create layering of graph, potentially using domain knowledge, longest path from root, algorithm Step 2 - Minimize crossings layer by layer Step 3 - Final assignment of x-coordinates, Routing of edges pros: Nice, readable top down flow - Relatively fast cons: Not suitable for graphs that don't have an intrinsic top down structure - Can be hard to implement

What are stream graphs?

Stream graphs are an extension of stacked bar charts where there is continuity at one axis. They are often used to indicate the popularity of a topic or artist over time, encoded by the bar heights.

Node Link layouts

Sugiyama layout: Layered approach where depth in graph mapped to one axis. Top down Force directed: Model layout constraints using physical forces. Edges are springs and nodes are repulsive paritcles

Example of text visualizations

Tag Clouds (visualizes word count) Word trees (Visualize word sequences) Topic Model Visualization (id topics from document) ThemeRiver (Thematic changes in document collections over time, is same as stream graph) PhraseNets (Pattern visualization)

Sources of sets

Tags • Queries • Subscriptions, votings, surveys • Probablistic events • Fuzzy clustering • Multi-label classifications

What is the impact of aspect ratio on line charts?

The aspect ratio of line charts affects how accurately we can judge the rate of change. The most accurate angle judgment happens around the 45-degree angle. Very steep or very flat angles are harder to compare accurately.

What is the expressiveness and effectiveness principle in visualization?

The expressiveness principle states that the visual encoding should express just the attributes in the data, but not more. The effectiveness principle states that the most effective channels should be used to encode the most important attributes in the data.

What is the challenge with the placement of ticks in charts?

The placement of ticks is a non-trivial problem as they need to be at round numbers, not too sparse and not too frequent. There are various approaches to handle this.

What is the role of trend lines in data visualization?

Trend lines are smooth versions of the original data that help in identifying the overall pattern or direction in the data.

Networks and trees

Used to describe communication patterns, telephone systems, computer networks Nodes and links and Graphs with hierarchical structure - Nodes referred to as parents and children

Set visualizations

Venn Diagrams Euler Diagrams Set'o'Gram Parallel Sets Radial Sets Bubble Sets Line Sets Kelp Diagrams

Tree layout

Visual structures that refer to use of connection and enclosure to encode relationships among cases Features: No crossing edges, reflects relationship among nodes, clean and non convoluted, heiarchial 4 approaches: Indentation, NodeLink, enclosure, layering

pie chart

a chart that shows the relationship of a part to a whole

Venn and euler diagrams represent

containment, intersection, exclusion Only small number of sets possible

Enclosure Diagrams

• Encode structure using spatial enclosure - Hierarchy is represented implicitly (by spatial arrangement) as opposed to explicitly (by link marks) Easier to spot large and small nodes, and to see entire tree

Summary of graphs

• General graphs - Node-link: familiar, but problematic for dense graphs - Adjacency matrices: abstract, hard to follow paths - Attribute-driven: not always possible • Trees - Indentation: simple, effective for small trees - Node-link and layered: look good but needs exponential space - Enclosure (treemaps): great for size related tasks but suffer in structure related tasks

Layered diagrams

• Recursive subdivision of space where Higher-level nodes get a larger layer area examples: Icicle trees, sunburst trees

Strength and limitations of matrix layout

• Strengths - Great for dense graphs - Visually scalable - Can spot clusters • Limitations - Abstract visualization - Hard to follow paths Can spot patterns in matrix

Principle types of tasks in networks and trees

- Attribute-based: related to the values associated with nodes or edges - Topology-based: related to the adjacency relationships in the graph

Reingold tilford layout

- Bottom-up recursive approach - For each parent, make sure every subtree is drawn - Pack subtrees as closely as possible - Center parent over subtrees

orthogonal, circular, nested layouts

- Great for UML diagrams - Algorithmically comple - Emphasizes ring topologies - Used in social network diagrams - Recursively apply layout algorithms - Great for graphs with hierarchical structure

How does a dense pixel display work?

Dense pixel displays use a single pixel for one particular item. The focus is not on individual attribute values, but rather on the overall structure of the data, identifying clusters in the data that can then be explored further.

Where does textual data comefrom and why visualize it

Documents like articles books and novels Text snipets like tweets, sms, tags ETC Visualize for understanding, summaries, clustering, quantify, correlate.

Mostif simplification

Extension of hybridVisualization, subnetworks with common patterns of nodes and links. Then we replace the common motifs by glyphs example clique motif, fan motif, connector motif

What are Generalized Pair Plots?

Generalized Pair Plots extend the scatterplot matrix concept by allowing different types of plots in each cell of the matrix. This allows for more flexibility in displaying relationships between variables, accommodating both quantitative and categorical data. For example, a cell could contain a scatter plot, box plot, or bar chart depending on the nature of the variables being compared.

Examples of tree layouts

Indentation (Items along vertical spaced rows, think file system. Need scrolling) Node link diagram (Reingold-tilford layouts, recursive approach) Enclosure diagrams (Treemaps, cushion t-map) Layered diagrams (Icicle trees, sunburst trees)

What is a multi-scale approach in line chart creation?

It is a method where interesting regions are identified at multiple scales. This can be done with a Fourier transformation, selecting aspect ratios as spikes in the Fourier transform in the power spectrum.

What is a misleading use of line charts?

It is misleading to use line charts when the horizontal axis is a categorical variable. This indicates a trend where there semantically cannot be one, violating the expressiveness principle.

What is the principle of 'banking to 45 degrees' in line charts?

It is the deliberate specification of the aspect ratio of a line chart such that the majority of the angles are close to 45 degrees. This enhances the readability and interpretation of the chart.

Tasks and their principle types

Localize Find nodes that fulfill property, A: edge with max weight, T:all adjacent nodes of given node Quantify: count a numerical property, A: give the number of all nodes, T: give the IN-degree of a node Order: Enumerate nodes/edges according to criterion, A: Sort all edges according to weight T: Traverse grap starting from given node

What are marks in the context of data visualization?

Marks are graphical elements or primitives that can be categorized according to the number of spatial dimensions. They include zero-dimensional marks like points, one-dimensional marks like lines, and two-dimensional marks like areas.

NodeTrix

Node link diagram combined with matrices. Node link shows overall graph structure Matrices show communities

Graph representations

Node-link Matrix Implicit

Graph layouts

Node-link Matrix Orthogonal Circular Nested NodeTrix Motif Tree Treemaps cone etc

What are parallel coordinates?

Parallel coordinates are a common way of visualizing high-dimensional geometry and analyzing multivariate data. This technique involves a set of parallel lines (axes) where each represents one attribute. A data point is represented as a polyline that intersects each axis at the corresponding attribute value.

Hyperbolic layout

Perform tree layout in hyperbolic geometry, then project the result onto the Euclidean plane Also computable in 3d, projected onto a shpere

Why should we avoid using pie charts?

Pie charts are often discouraged because they can be misleading and difficult to interpret accurately. The human eye struggles to compare the size of angles and areas in a pie chart. Instead, alternatives like bar charts or normalized stacked bar charts are recommended as they use length or position, which are more accurately perceived by the eye.

What are radar plots?

Radar plots, also known as spider or star plots, are graphical methods of displaying multivariate data in the form of a two-dimensional chart. Each variable is represented on a separate axis that starts from the center point and data points are connected to form a polygon. They are often used for comparing profiles across multiple entities.

Treemaps and cushion treemaps

Recursively fill space based on a size metric for nodes Enclosure signifies hierarchy squares, circles, triangles etc and Treemaps where area looks like cushons, shading tells more about the topology. • Visual encoding:show nesting/topological structure more clearly with shading cues

Scatter Plot Graph

A graph of plotted points that show the relationship between two sets of data that do not depend on each other where the line is not connected. (Ex. height versus weight)

What is a heatmap?

A heatmap is a graphical representation of data where individual values contained in a matrix are represented as colors. It's often used to visualize large datasets and show patterns, correlations, and clusters in the data.

What is a scatterplot matrix?

A scatterplot matrix is a grid of scatter plots representing all possible pairs of variables in a dataset. It allows for the visualization of multi-dimensional data, where each scatter plot in the matrix shows the relationship between two variables.

What is a stacked bar chart?

A stacked bar chart is a common approach to encode more than one key. It has one quantitative attribute encoded in the length and two categorical attributes. One of the categorical attributes is along the horizontal axis, and the other is encoded using color.

Overlays in set visualization

Augmenting set memperships over, tables and lists, maps Region based (bubble sets) Line based (Line sets) Glyph based (Kelp diagrams)

What is the purpose of bar charts and line charts in data visualization?

Bar charts and line charts are great for comparing values and looking up values. They are particularly useful when the horizontal axis is a categorical variable.

Different overlay set visualizations

Bubble Sets: Display additional membership relationships over existing visualizations like scatterplot Line Sets • Reduces representation of set membership as lines to avoid visual clutter can show it over a map for ex Kelp Diagrams • Use diagrammatic layout algorithm to avoid conveying misleading membership relationships

What is the difference between categorical, ordinal, and quantitative attributes in tabular data?

Categorical attributes have no ordering relationship and belong to a particular group. Ordinal attributes have an ordering relationship but cannot be used for arithmetic operations. Quantitative attributes can be used for arithmetic operations and have a clear ordering.

What are channels in data visualization?

Channels control the appearance of marks. They include position channels, shape channels, size channels like lengths and areas, color channels, and even channels like tilt or volume.

What are some common strategies to visualize tabular data?

Common strategies include scatter plots, bar charts, dot plots, and line charts. Each has its own strengths and limitations in terms of scalability and the number of attributes that can be displayed.

What types of data are commonly visualized in information visualization?

Common types of data include tables containing items and their attribute values, networks, trees, set value data like text, and image sets.