Last Lecture Flashcards
Classifications of users
- Familiarity with domain
- Familiarity with task
- Familiarity with data
- Familiarity with the visualization technique
- Familiarity with the visualization environment
Ideal evaluation using human subjects
Want range of characteristics of participants to be as similar as possible to the intended audience
Different varieties of data
- Type
- Size
- Dimensionality
- Number of parameters (univariate, multivariate
- Structure (table vs. hierarchical)
- Range
- Distribution
- Real vs. Synthetic data
Visualization characteristics
- Computational performance
- Memory performance
- Data limitations
- Degree of occlusion
- Degree of complexity
- Degree of usability
- Degree of accuracy
Tests for evaluating visualizations
- Usability test: observing users perform tasks in controlled environment over short period of time
- Field tests: performed in natural environment of typical user, last for weeks or months
- Case studies and use cases
Five E’s of usability test
- effective
- efficient
- engaging
- error tolerant
- easy to learn
Steps in benchmarking
- Formulate a hypothesis
- Design the experiments (vary only in a single attribute at a time)
- Execute the experiments (each participant get similar instruction)
- Analyze the results and validate the hypothesis (supported, refuted or insufficient evidence; need to be statistically significant)
Outlier Detection and Measurement Experiments
Stage 1, 2, 3, and 4
Stage 1: develop quantifiable definition for an outlier and create algorithm capable of labeling data points appropriately
Stage 2: Data sets analysed that contained outliers according to definition in Stage 1
Stage 3: Subjects need to:
- Determine if image contains outliers
- Identify points believed to be outliers
- Estimate degree of separation on 5-point scale
Stage 4: usefulness of each visualization method was tested
