Data Abstraction – Lecture 03
Analysis framework: Four levels, three questions
Domain sitation
- Who are the target user?
Abstraction
- Translate from specifics of domain to vocabulary of visualization
- How is it shown?
- Efficient computation
Nest model
- Downstream: cascading effects
- Upstream: iterative refinement
What does data mean?
Semantics: real-world meaning of the data
- Column names help with semantics
- May also include rules about data
- Useful for asking good question about the data
Data types: structural or mathematical interpretation of data
- Item: individual entity, discrete
- a.k.a independent variable
- Attribute: some specific property that is measured, observed, or logged
- a.k.a dependent variable
- Links
- Express relationship between two items
- Positions
- Spatial data
- Grids
- we can infer some of this information
Dataset types – see lecture 03…..
Tables
- Flat table
- one item per row
- each column is an attribute
- each cell holds value for item-attribute pair
- unique key (could be implicit)
- Multidimensional table
- nodes (vertices) connected by links (edges)
- node: synonym for item in the context of networks
- link: a relation between two nodes
- tree is a special case: no cycles
- Grids: specify how data is sampled
- Positions: a location in space
Atrribute Types
- Categorical
- Ordered
- Ordinal
- Quantitative
- time
Ordering Direction
- Sequential
- Diverging
- Cyclic
Dataset Availability
- Static
- Dynamic
Data abstraction: Three operations
Data abstraction: Translate from domain-specific language to generic visualisation language
- Identify dataset types, attribute types
- Identify cardinality
- how many items in the dataset?
- what is the cardinality of each attribute?
- number of levels of categorical data
- range for quantitative data
- Consider whether to transform data
- guided by understanding of task
Task Abstraction – Lecture 04
From domain to abstraction
- Domain characterisation: details of application domain
- group of users, target domain, their questions & data
- domain questions / problems
- break down into simpler abstract tasks
- Abstraction: data & task
- map what and why into generalised terms
- identify tasks that users wish to perform, or already do
- find data types that will support those tasks
- possibly transform / derive if need be
Task abstraction: Actions and targets
Very high-level pattern
Actions
- analyse
- high-level choices
- visualisation for consumption
- discover new knowledge
- generate new hypothesis or verify existing one
- designer doesn’t know what users need to see
- present known information
- presenter already knows what the data says
- wants to communicate this to an audience
- enjoy
- similar to discover, but without concrete goals
- discover new knowledge
- visualisation for production
- annotate
- record
- persist visualisations for historical record
- derive
- create new data
- create derived attributes
- crucial design choice
- search
- find a known/unknown item
- what does the user know?
- lookup: target known, laocation known
- locate: target known, location unknown
- browse: target unknown, location known
- explore: target unknown, location unknown
- query
- what is being acted on
- All data
- trends
- outliers
- features
- Attributes
- One
- Distribution
- extremes
- Distribution
- Many
- Dependency
- Correlation
- Similarity
- One
- Network Data
- Topology
- Paths
- Topology
- Spatial Data
- discover distribution
- compare trends
- locate outliers
- browse topology
Marks and Channels – Lecture 05
Visual encoding of information:
- Marks: the basic graphical elements in a visualisation
- Channels: way to control the appearance of the marks based on attributes
Marks as constraints
- geometric primitives
- classified by dimensionality
- 0D: points
- 1D: Lines
- 2D: Areas
- ……
- marks for links
- control appearance of marks
- type and amount of information that can be conveyed to human perceptual system
- classified
- match channel type to data type
- expressiveness principle: all data from the dataset and nothing more should be show
- some channels are better than others
- effectiveness priciple: the most important attributes should be the most salient
- saliency: how notceable something is
- how do the channels we have discussed measure up?
- accuracy
- how precisely can we tell the difference between encoded items
- discriminability
- how many unique steps can we perceive?
- separability
- is our ability to use this channel affected by another one?
- poppit
- magnitude channels: ordered attributes
- position on common scale
- position on unaligned scale
- length
- tilt / angle
- area
- depth
- color luminance
- color saturation
- curvature
- volume
- identity channels: categorical attributes
- spatial region
- color hue
- motion
- shape
- Accuracy / discriminability / separability / pop-out of visual channels
- Relative vs absolute judgements of visual perception system