DAVI Visualizing Time-oriented Data

Lecture Overview

• Today’s Focus:
  • Visualization of temporal data (time visualization)
  • Discussion of two visualization critiques
• Upcoming Sessions:
  • Design exercises in preparation for an industry customer visit

Lecture Recap (Mentee Questions and Answers)

1. Parallel Coordinates Pattern
   • Question: In parallel coordinates, what does a star-shaped pattern indicate?
   • 
   • Answer: A negative correlation
   • Explanation: High values on one axis connected to low values on another axis form a star shape, indicating negative correlation.
2. Chart Identification
   • Question: What type of chart is this?
   • 
   • Answer: A Marimekko chart (also acceptable: Mosaic plot)
   • Explanation: Marimekko charts are used for quantitative data, similar to mosaic plots which are for categorical data. They bin value ranges and display proportions.
3. Use of Parallel Sets
   • Question: Parallel sets are used for what type of data?
   • Answer: Qualitative data (categorical data)
   • Explanation: Parallel sets visualize relationships between multiple categorical variables, serving as an alternative to mosaic plots.
4. Trellis Plot Characteristics
   • Question: What is true about a trellis plot?
   • Answer: It subdivides the dataset
   • Explanation: Trellis plots divide data into subsets and display them in a grid, facilitating comparison across different categories.
5. Van Wijk’s Model Focus
   • Question: Van Wijk’s model of visualization details which aspect of the visualization pipeline?
   • 
   • 
   • Answer: The use of the visualization
   • Explanation: Van Wijk’s model expands on how users interact with and derive knowledge from visualizations, focusing on the cognitive processes involved.

---

Principles of Time Visualization

Understanding Time in Visualization

• Timescales:
  • Ordinal: Events are in order but without quantifiable intervals.
    • 
  • Discrete: Specific, separate time points (e.g., timestamps).
    • 
  • Continuous: Time is measured on a continuous scale, allowing for interpolation.
    • 
• Time Scope:
  • Time Points: Values valid at specific moments.
    • 
  • Time Intervals: Values valid over a duration until the next measurement.
    • 
• Time Arrangement:
  • Linear Time: Chronological progression (past to future).
    • 
  • Cyclic Time: Repeating patterns (e.g., seasons, daily cycles).
    • 
  • Branching Time: Forking paths in time (e.g., version control systems: Git).
    • 
• Time Granularity(粒度):
  • 
  • Different levels of detail (seconds, minutes, days, months).
  • Multiple granularities can coexist, forming hierarchies like calendars.
  • it may make sense to have a visualization that supports this view on multiple granularity at the same time.

Implications for Visualization

• The combination of timescale, time scope, and time arrangement influences the choice of visualization techniques.
  • 
• Examples:
  • Continuous Timescale with Time Points (Linear): Line charts with time on the x-axis.
  • Discrete Timescale with Time Intervals (Linear): Bar charts showing durations.
  • Continuous Timescale with Intervals (Gantt Charts): Visualizing tasks or events over time.
• Can a continuous time scale (as on the left) and interval time scopes (as on the right) ever co-occur? If not, why? If so, how could a visualization for them look like?
  • 

---

Tasks on Time-Oriented Data Visualization

• Standard Tasks:

  • Reading specific values
  • Identifying extremes or outliers
  • Finding clusters or patterns
  • Gauging distributions

• Time-Specific Tasks:

  • Analyzing data

    behavior over time

    • In Fluctuating Data (Heterogeneous -> 各種各樣的；混雜的):

      Look for regularities.

      • Trends: General directions in data over time.
      • Periodicities: Repeating patterns at regular intervals.
      • Sequences: Specific orders of events.
      • 

    • In Homogeneous(同类事物（或人）组成的；同类的；相似) Data (Predictable):

      Look for deviations.

      • Outliers: Data points that differ significantly.
      • Fluctuations: Irregular changes or noise.
      • Irregularities: Unexpected sequences or missing patterns.
      • 
      • 

---

Mapping Time to Visual Attributes

Using Visual Channels (Based on Mackinlay’s Ranking)

• Position: Most effective for representing time (e.g., x-axis in line charts).

• Length: Represents durations or intervals.

• Angle: Useful for cyclic time (e.g., clock faces, circular plots).

• Connection: Shows progression or sequence (e.g., connected scatter plots).

• Size and Texture: Can indicate temporal progression but less commonly used.

• Color:

  • Luminance (Brightness): Can represent time progression.

  • Hue: Generally not recommended for time due to categorical perception.

    • Ps: Hue distinguishing it as red, blue, green defined by its position on the color wheel or by its dominant wavelength

  • we usually don’t use color if we indicate um, or if we equate temporal a temporal attribute with a quantitative

    • 颜色的顺序性有限：人类对颜色的感知并非线性，且颜色之间缺乏自然的顺序关系。因此，使用颜色表示时间序列可能导致对时间进程的误解。

      颜色感知的主观性：不同的人对颜色的感知可能存在差异，受文化、背景和个人经验等因素影响。这使得颜色在表示精确的时间信息时不够可靠。

      色盲和视觉障碍：约有8%的男性和0.5%的女性存在某种形式的色盲。使用颜色编码时间信息可能导致这些人群无法准确解读数据。

      颜色数量的限制：人类对颜色的区分能力有限，通常只能有效地区分有限数量的颜色。对于需要表示大量时间点的数据，颜色编码可能无法满足需求。

• Shape: Not effective for time; shape is an identity channel without inherent order.

Mapping Techniques

• Time to Space (Small Multiples):
  • 
  • Displaying multiple snapshots side by side for comparison.
  • Limitation: Space constraints with many time points.
• Time to Display Time (Animation):
  • 
  • Showing changes over time through motion.
  • Limitation: Difficult to compare different time points simultaneously.
• Combining Both:
  • 
  • Example: Gapminder uses animation with trails to show progression and maintain context.

---

Visualization Techniques for Time Data

Line Charts

• Standard method for continuous time-series data.
• Time on the x-axis, value on the y-axis.
• 

Horizon Graphs

• Purpose: Efficiently display multiple time series in limited space.
• 
• Method:
  • Convert data into bands using color to represent value ranges.
  • Flip negative values to the positive side, using color to distinguish them.
  • Layer bands to reduce vertical space required.

Calendar Visualizations

• Show data across different granularities (days, weeks, months).
• Useful for identifying patterns like daily or weekly trends.

Gantt Charts

• Visualize tasks or events over intervals.
• Time on the x-axis, tasks or events on the y-axis.
• 

Triangular Model

• Alternative to Gantt Charts for Large Datasets
• Method:
  • Plot intervals as points in a scatterplot.
  • X-axis: Start time
  • Y-axis: Duration
• Advantages:
  • Handles large numbers of intervals efficiently.
  • Helps identify patterns like overlapping events.
• 其他的表现形式

Arc Diagrams

• Visualize sequences and repetitions over time.
• Arcs connect similar or repeating events.

Outflow Graph

• Purpose:Analyze event-based data. This is a graph that visualizes a sequence of events (or a sequence of behaviors), often referred to as an “Outflow Graph” or “Sankey-like” visualization, and is used to express transfer relationships between different events.

• Horizontal (Width): The time it takes from the beginning of an event to the next event (or relative time span).
• Vertical (Height): indicates the number of times/frequency an event occurs. The higher the bar, the more often this event occurs in all sequences.
• Color: Indicator that usually maps to some kind of “Outcome” or “result” (legend goes from red to green, values from 0.0 to 1.0). The different colors represent the “positive” or “negative” outcome of the event, or other characteristic dimensions.

Time Maps

• Purpose: Analyze event-based data.
• Method:
  • X-axis: Time since the previous event.
  • Y-axis: Time until the next event.
• Application:
  • Identifying patterns in event sequences.
  • Distinguishing human activity from automated patterns (e.g., detecting bots on Twitter).

---

Interaction Techniques

Multi-Level Zoom (Stack Zooming)

• Concept:
  • Users select regions to zoom into, creating a hierarchical stack of detailed views.
• Benefits:
  • Maintains context with the overall dataset.
  • Allows exploration at multiple levels of detail.

ChronoLens

• Interactive Tool for Exploratory Analysis
• Features:
  • Create “lenses” to focus on specific data regions.
  • Apply transformations and computations (e.g., filters, aggregations).
  • Build analysis pipelines by chaining lenses.
• Advantages:
  • Facilitates deep exploration without losing context.
  • Supports dynamic adjustments and real-time feedback.

---

Visualization Critiques

1. America’s Religious Landscape Pie Chart

• Issues Identified:
  • Mixing Categories: Combines religion and ethnicity inconsistently.（EXPRESSIVENESS）
  • Incomplete Data: Does not break down unaffiliated groups by ethnicity.（EXPRESSIVENESS）
  • Percentages Don’t Sum to 100%: Likely due to rounding errors.（EXPRESSIVENESS）
  • Ineffective Use of Pie Chart:
    • Pie charts are not ideal for comparing complex hierarchical data. （EFFECTIVENESS）
  • Color Choices:
    • Not colorblind-friendly (issues with green and orange hues).（EFFECTIVENESS）
  • Ordering:
    • Segments are not consistently ordered by size or category.
• Suggestions for Improvement:
  • Use a Sunburst Chart:
    • Displays hierarchical data effectively.
    • Inner rings represent higher-level categories (religion), outer rings show subcategories (ethnicity).
      • 
  • Separate Charts:
    • Provide individual charts for religion and ethnicity to avoid confusion.
  • Correct Percentages:
    • Ensure percentages add up to 100% by adjusting rounding or displaying decimals.
  • Improve Color Scheme:
    • Use colorblind-friendly palettes.
    • Avoid implying order where none exists.

2. Bivariate Map of Wisconsin

• Issues Identified:
  • Lack of Normalization:
    • Bubble sizes represent the absolute number of+ high school graduates, not accounting for population differences.
    • Larger counties appear to have more graduates simply due to higher population.
  • Incorrect Bubble Scaling:
    • Bubbles are scaled by diameter instead of area, exaggerating differences. -> lie factor
  • Inconsistent Binning:
    • Household income categories in the choropleth map are unevenly binned without clear reasoning.
  • Poor Bubble Placement:
    • Overlapping bubbles and misalignment with county boundaries.
• Suggestions for Improvement:
  • Normalize Data:
    • Use percentages (e.g., proportion of residents with high school degrees) to allow meaningful comparisons.
  • Correct Bubble Scaling:
    • Scale bubbles based on area to accurately represent data.
  • Consistent Binning:
    • Use equal intervals or quantiles for income categories.
    • Provide clear explanations for chosen binning methods.
  • Improve Visual Clarity:
    • Adjust bubble placement to reduce overlap.
    • Consider alternative visualization techniques, such as proportional symbols or graduated color maps.

---

Key Takeaways

• Data and Task Alignment:
  • Choose visualization techniques that align with the data characteristics and the intended analytical tasks.
• Avoid Common Pitfalls:
  • Be cautious with visual encoding (e.g., scaling symbols correctly, using appropriate color channels).
  • Ensure data is represented accurately and clearly.
• Interactivity Enhances Exploration:
  • Interactive tools like ChronoLens can significantly aid in understanding complex time-series data.
• Consider Audience and Accessibility:
  • Use colorblind-friendly palettes.
  • Provide clear legends and labels.
  • Ensure that visualizations are interpretable without requiring excessive effort.

---

Next Lecture Preview

• Topic:

  Geospatial Data Visualization

  • Map projections
  • Techniques for displaying data on maps

• Preparation:

  • Review the relevant textbook chapter on geospatial visualization.

---

References:

• Books:
  • “Visualization of Time-Oriented Data” by Aigner, Miksch, Schumann, and Tominski (Springer)
  • “Visualizing Time” by E. G. Tufte
• Websites:
  • TimeViz Browser: http://www.timeviz.net
• Visualization Tools:
  • Gapminder: https://www.gapminder.org/tools/

---

Additional Notes:

• Importance of Time Visualization:
  • Time is a critical dimension in many datasets.
  • Effective visualization of temporal data can reveal trends, patterns, and anomalies that are not apparent in static data.
• Multi-Granularity Analysis:
  • Visualizations that support multiple time granularities allow users to zoom in and out of data, observing both high-level trends and detailed patterns.