Skip to content

lesson-4-data-visualization.md

Latest commit

 

History

History
87 lines (85 loc) · 2.95 KB

lesson-4-data-visualization.md

File metadata and controls

87 lines (85 loc) · 2.95 KB
Raw

Lesson 4 - Data Visualization

  • Information Visulization

    • Effective communication of complex quantitative ideas
      • Clarify
      • Precision
      • Efficiency
    • Helps you notice things about data (correlations, trends)
    • Highlight aspects of data, or "tell a story"

  • Communicating findings

    • Don's advice
      1. "Craft a narrative"
      2. Know your audience
        • Technically minded?
        • People you want to recruit
        • Customers
    • Rishi's advice
      • Make is easily understandable but have math/stat rigour

  • Visual Encodings

    • Position
      • Example: Positial data on chart
    • Length
      • longer the bar, greater the value
      • Example: Bar chart
    • Angle
      • Example: pie chart
        • the higher the degrees, the bigger the "slice"
        • generally avoid if showing very small differences
    • Direction
      • Has similar problems it angle. Can be hard to see differences.
    • Shape
      • use to differentiate types of data (different teams, districts)
    • Volume
      • representing data with size
    • Colour
      • hue
        • categorical data
      • saturation
        • intensity of colour for a hue

  • Plotting with ggplot

    1. Create plot
    > ggplot(data, aes(x_var, y_var))

      * ```data``` == dataframe
  * ```aes(x_var, y_var)```
    1. Represent data with geometic objects
      • geom_point() - change property of points
      • geom_line() - change property of lines
    2. Add labels *ggtitle('Title') - to title plot
      • xlab('X Label') - label x
    • Example:
    >> print ggplot(df, aes(x="yearID", y="HR")) + geom_point(colour="red") + geom_line()

  • Data types

    • Numeric data
      • A measurement (height, weight) or count (HR or hits)
    • Discrete and continuous
      • Discrete: Can only have whole number values
      • Continuous: any number within range
    • Categorical data
      • Represent characteristics (eg position, team, hometown, handedness)
    • Ordinal data
      • Categories with some order or ranking
      • Movie: between 1 star and 5 stars
      • Same as categorical but ordered
    • Timeseries
      • Collection of numbers collected in intervals over time

  • Scale

    • Scale must be inconsistent

  • Visualizing Time Series Data

    • Scatterplot without lines can make it hard to view trends
    • Linechart may focus on year-to-year variability instead of overall trends
    • LOESS curve can capture long term trends

  • Multivariate data

    • Use scale to show where additional events occured with the base event
    • Double up on visual queues: area and colour

  • Rishraj's advice

    • Learn tools well
    • Use them in the correct way

  • Don's advice

    • Difference between good and mediocre ds
      • Feature-selection process
      • Learn as many mathematical tools as possible