Module 6 - Complex Idioms Flashcards
Why Maps? (V1)
- Understanding spatial relationships
- Familiarity
* People know where something is on a map
(assuming they are familiar with the region) - Maps act as an index from spatial to semantic information and vice-versa.
Choropleth map (V1)
Data = 1 Quantitative attribute w/ geographic geometry
Mark = Geometric Area
Channel = Color
Tasks = Spatial relationship
*misleading (size of objects depend on geography, not the attribute), cannot compare well
-> resolve by using glyphs
mark = dot/point
channel = size
Cartogram (V1)
a map that distorts such that area represents size/quantity correctly
*distorting and accuracy at the cost familiarity
Dot Map (V1)
Introduce a point that represents observations, and scatter across map
Density Map (V1)
Turn discrete data into continuous data (using KDE) and plot as a heatmap
Can also be 2d, with transparency representing bins
Topographic Map (V1)
Data = Scalar Spatial Field (quantitative continuous data, can interpolate between points) w/ geographic geometry
Mark = Lines,
Channel = Shape, Position Color
Tasks = Spatial Relationship
Caution with Maps (V1)
A dataset may contain geographical information
* Yet creating a geographical visualization may not be
relevant.
* Ask: “does spatial arrangement matter for my task?”
* Position most effective visual channel do not waste
it, if not relevant.
Absolute vs Relative
Network (V2)
Describe the relationship between objects
represented as tuples of verticies and edges
- Undirected OR Directed
Tree (V2)
Special type of network that has hierachy
- a graph without cycles (acyclic) and has one root
- nodes have 0 or more children
Types of Static Networks (Design) (V2)
Types (Design choices for arrangement):
1. Node-link diagram (both)
2. Adjacency Matrix (both)
3. Enclosure (does not represent network, only tree)
Structure is important, shows areas of interest
Types of Network Structure (V2)
- Radial
- hides a lot of info, structure depends on ordering of nodes - Arc
- edges move clockwise. top are left to right, bottom right to left
- linearise nodes! (sort/ordering)
Layout/Embedding (V2)
Positioning of nodes
Things to do when Visualising a Static Network (V2)
- Maximise Readability & aesthetics
* Equal edge length
* Minimize crossings
* Non-overlapping nodes
* High-degree nodes should have a central position
* Symmetry should be maximized
* Communities should be clearly visible - Optimise structure using node force directed algorithms
Issues caused by Large Networks, Solution? (V3)
Methods used to visualise static network do not apply, as there is too much data -> hairball visualisation
Use other attributes from data such as Hierachy (from data OR computed) and node+edges Force directed algorithms to give structure
Visual Adjacency Matrix for Large Networks (V3)
- More scalable compared to node link diagram
Use color (transparency) to denote weight of edges - Maximise edge visibility
*no crossings, cannot find paths easily
*ordering of nodes is important
How to get good ordering for Visual Adjacency Matrix, for Large Networks (V3)
Additional Hierachal Data (from data OR computed) (hieracheral clustering)
Which (design) choice of networks is used for Trees? (V4)
- Node Link diagram
- Enclosure
(can combine both methods into one visualisation)
adjacency matrix is hardly used due to the structure of the tree, data will be very sparse
Nodelink Diagrams for Trees (V4)
standard node-link diagram wastes a lot of space, instead use radial layout (root node in middle)
if tree is very big, use hyperbolic technique
Enclosure Diagrams for Trees (V4)
- Venn Diagram,
- Treemap
- Nested Treemap
Treemap VS Nodelink (V4)
Treemap:
+ scalability is milions of items, better then node link. limit is screen resolution
+ Good usage of Space
- Hierachy/order is implicit (not shown directly)
Node-link
+ intuitive
+ good at exposing structure of information
- empty space
Treemap Space Filling Techniques (V4)
Slice and Dice Layout:
* leads to very long/elongated rectangles (due to switching between vertical and horizontal cuts),
use Squarified Layout instead:
* however, comparison between dynamic hierachy is impossible
Dynamic Networks (V5)
Networks that change over time
- edge is represented by vertices AND time-stampP
Properties of Dynamic Networks (V5)
Structural Properties:
- communities
- motifs
Temporal Properties:
- Trends
- Anomalies
- Periodicity
- Temporal Shifts
Understanding Dynamic Networks (V5)
Discovering/Exploring states
* Characterizing the evolution of
the network
* Stable states, recurring states,
outlier states
* Transitions between them
Methods to visualise Dynamic Networks (V5)
- Animation (mapping time to time)
- Small Multiples (mapping time in intervals)
- Integrated approach (mapping time to space)
Animation to Dynamic Networks (V5)
for each time stamp, compute layout -> show changes
+ easy to implement
+ easy to spot big changes
+ applicable to all methods that a time stamp can be visualised
- need to focus on many moving items simultaneously
- keep track of multiple changes over time (relying on user memory)
- change blindess (cannot see small changes)
use:
- preservation of the mental map (keep variation between layouts as small as possible)
- add timeline control
Small Multiples to Dynamic Networks (V5)
Juxtaposed visualisation using filmstrip or a grid layout
+ Can be used for any dynamic network
- Decide on number of multiples used (too many = hard to spot changes, too little = can’t see change)
Integrated Approach (Mapping time to space) to Dynamic Networks (V5)
Provide static overview of entire time span of network in one visualisation
+ Complete overview
+ Global patterns are easily identified
- Specialised Visual encoding, difficult to interpret for non experts
- Restricted to specialised type of network
Examples of Integrated Approach (Mapping time to space) (V5)
Clusters over time (Individual nodes and edges not visible anymore)
Massive Sequence View (ordering is important)
Parallel Edge Splatting (needs interaction)
Idioms for Time-Series Data
Line Chart, Streamgraph
Connected Scatterplot:
same as normal scatterplot, but have multiple points connected using directed lines to show temporal ordering
Gantt Chart:
Data = 1 Categorical Attribute (Key), 2 Quantitative Attributes
Mark = Line
Channel = Horizontal position (start position), Length (duration)
Tasks = Temporal Overlaps, dependencies between categories
