Skip to main content

Introduction to Data Science: A Low-Code Project-Based Approach

Hannah Kurzweil

Section Visualization Fundamentals

Subsection The Power of Data Visualization

Data visualization serves several crucial functions in the data science process:
Exploration
Visualizations help us discover patterns, relationships, and outliers that might not be apparent in raw data or summary statistics.
Analysis
Visual representations allow us to compare groups, identify trends, and test hypotheses about our data.
Communication
Well-designed visualizations make complex findings accessible to others, supporting decision-making and action.
Engagement
Visual elements capture attention and make data more engaging and memorable than text or numbers alone.
The power of visualization is illustrated by Anscombe’s Quartet
 15 
https://en.wikipedia.org/wiki/Anscombe%27s_quartet
, a set of four datasets with nearly identical statistical properties but dramatically different visual patterns:
Four scatter plots showing datasets with identical summary statistics but very different patterns: a linear relationship, a curved relationship, a linear relationship with an outlier, and a vertical line with an outlier.
Figure 85. Anscombe’s Quartet
Each of the four datasets in Anscombe’s Quartet has the same mean, variance, correlation, and linear regression line. If we relied only on these summary statistics, we’d miss the crucial differences between these datasets. Visualization reveals what statistics alone can hide.

Checkpoint 86. Visualization Purpose.

    Which of the following is NOT a primary purpose of data visualization?
  • To discover patterns and relationships in data
  • This is a primary purpose of data visualization - helping analysts discover patterns, trends, and relationships that might not be apparent in raw data.
  • To communicate insights to diverse audiences
  • This is a primary purpose of data visualization - making complex findings accessible and understandable to various stakeholders.
  • To increase the volume of data that can be presented
  • Correct! While visualizations can represent large datasets, the primary goal is not to maximize the volume of data presented. In fact, effective visualization often involves carefully selecting what data to present to avoid overwhelming the viewer.
  • To identify outliers and unusual patterns
  • This is a primary purpose of data visualization - making outliers and anomalies in the data immediately visible for further investigation.

Subsection Choosing the Right Visualization

Different types of data and questions call for different visualization approaches. Selecting the right chart type is essential for effectively representing your data.
Here are some common visualization types and their appropriate uses:
Bar Charts/Column Charts
Best for comparing values across discrete categories.
  • Vertical bars (columns): Good for comparing values across categories
  • Horizontal bars: Useful when category names are long or there are many categories
  • Grouped bars: For comparing multiple measurements across categories
  • Stacked bars: For showing composition within categories
Line Charts
Best for showing trends over time or continuous variables.
  • Single line: Shows trend in one variable
  • Multiple lines: Compares trends across groups
  • Area charts: Emphasize magnitude along with trend
Scatter Plots
Best for examining relationships between two numerical variables.
  • Simple scatter: Shows correlation between two variables
  • Bubble charts: Add a third dimension through point size
  • Scatter with groups: Compare relationships across categories
Pie Charts
Best for showing parts of a whole (composition) when there are few categories.
  • Use only when the sum is meaningful and categories are few (preferably 5 or fewer)
  • Consider alternatives like stacked bars for more effective comparison
Histograms
Best for showing the distribution of a single numerical variable.
  • Shows frequency of values falling within specific ranges (bins)
  • Reveals shape of distribution (normal, skewed, bimodal, etc.)
Box Plots
Best for showing distribution summary and comparing distributions.
  • Shows median, quartiles, range, and outliers
  • Excellent for comparing distributions across groups
Heat Maps
Best for showing patterns across two categorical dimensions.
  • Uses color intensity to represent a third variable
  • Good for large datasets with complex patterns
Maps
Best for showing spatial patterns and geographic distributions.
  • Choropleth maps: Color regions based on data values
  • Dot maps: Show individual locations or events
Examples of various chart types including bar chart, line chart, scatter plot, pie chart, histogram, box plot, heat map, and map.
Figure 87. Common Chart Types from Medium.com
 16 
https://medium.com/@ThisIsShubhangi/data-visualization-choosing-the-right-chart-matters-d4f6b75f8ae0
.
When choosing a visualization type, consider:
  • What question are you trying to answer?
  • What types of variables are you working with (categorical, numerical, time-based)?
  • What relationship or pattern are you trying to show (comparison, composition, distribution, relationship)?
  • Who is your audience and what is their level of data literacy?

Checkpoint 88. Matching Chart Types to Questions.

Activity 25. Chart Selection for Your Dataset.

In this activity, you’ll identify appropriate visualization types for different aspects of your dataset.
(a)
Write down at least three specific questions you want to answer with visualizations of your dataset.
(b)
For each question, identify:
  • The variables involved (including their types: categorical, numerical, time-based)
  • The most appropriate visualization type to answer the question
  • Why this visualization type is the best choice for your specific question
(c)
Share your choices with a classmate and provide feedback on each other’s selections.

Subsection Principles of Effective Visualization

Creating effective visualizations requires more than just selecting the right chart type. Well-designed visualizations follow these key principles:
Show the Data
The primary purpose of a visualization is to clearly present the data. Avoid unnecessary decorations or "chart junk" that distracts from the data.
Reduce Cognitive Load
Make it as easy as possible for viewers to extract insights. Minimize the mental effort required to interpret the visualization.
Focus Attention
Use visual elements like color, size, and position to direct attention to the most important aspects of the data.
Provide Context
Include reference points, comparisons, or baselines that help viewers understand the significance of the data.
Be Truthful
Represent the data accurately without distortion. Ensure that visual encodings are proportional to the data values.
Specific design principles include:
  • Clear labeling: Include descriptive titles, axis labels, and legends
  • Thoughtful color use: Choose accessible color schemes and use color consistently and purposefully
  • Appropriate scales: Use scales that accurately represent the data, generally starting at zero for bar charts
  • Data-ink ratio: Maximize the ratio of "ink" used to display data versus decorative elements
  • Simplicity: Include only what’s necessary, removing redundant or distracting elements
  • Consistency: Use consistent formats, scales, and styles across related visualizations
The Economist wrote a great article called, Mistakes, We’ve Drawn a Few
 17 
https://medium.economist.com/mistakes-weve-drawn-a-few-8cdd8a42d368
that I would highly recommend reading to see several ways visuals are created that are misleading and better practices to follow when creating visuals from your data. Below you can see one of the graphs from that article, where the bad graph attempts to create a perfect correlation between dog weights and neck size by creating the y-axes disproportionally.
Two versions of the same chart: one with clear labeling, appropriate scale, and thoughtful design; the other with misleading scale, excessive decoration, and unclear labeling.
Figure 89. Good vs. Problematic Visualization

Example 90. Visualizing Community Health Data.

For our Community Health dataset, effective visualizations might include:
  • A scatter plot showing the relationship between air quality index and asthma rates across neighborhoods, with points colored by income level to reveal potential socioeconomic patterns
  • A grouped bar chart comparing multiple health metrics (asthma, obesity, heart disease) across different regions of the city, allowing for direct comparison of health outcomes
  • A choropleth map showing the spatial distribution of environmental risk factors across the city, revealing geographic patterns and potential hotspots
  • Box plots comparing the distribution of health outcomes across neighborhoods with different levels of green space access, showing both central tendency and variation
  • A connected scatter plot showing how the relationship between environmental factors and health outcomes has changed over time

Checkpoint 91. Steps for Creating Effective Visualizations.

Arrange the following steps in a logical order for creating an effective data visualization.
Hint.
Think about how you would first plan what you want to show before deciding how to show it.

Activity 26. Visualization Critique.

In this activity, you’ll analyze and critique existing visualizations using the principles we’ve discussed.
(a)
Find three different data visualizations from news articles, reports, or websites. Try to include at least one visualization that you think is particularly effective and one that could be improved.
(b)
For each visualization, evaluate:
  • Is the chart type appropriate for the data and question?
  • How well does it follow the principles of effective visualization?
  • What works well about this visualization?
  • How could it be improved?
(c)
For the visualization that needs improvement, sketch or describe how you would redesign it to be more effective.

Subsection Creating Visualizations in CODAP

CODAP offers a variety of tools for creating interactive visualizations. Let’s explore how to use these tools effectively.
Key visualization features in CODAP include:
  • Graph tool: Create scatter plots, line plots, and dot plots by dragging attributes to axes
  • Plot type menu: Change between dot plots, box plots, histograms, and other display types
  • Axis options: Adjust scales, binning, and other axis properties
  • Grouping: Split data into categories using the ruler tool
  • Legend attributes: Add color, size, or shape encoding by dragging attributes to the legend area
  • Connecting lines: Show connections between related points
  • Map tool: Display geographic data on interactive maps
Screenshot showing CODAP’s interface for creating and customizing visualizations, with labeled tools for graphs, plot types, axis options, and other features.
Figure 92. CODAP Visualization Tools
Steps for creating effective visualizations in CODAP:
  1. Open the Graph or Map tool from the main toolbar
  2. Drag attributes from your dataset to the axes or legend area
  3. Select the appropriate plot type from the menu
  4. Adjust axis scales, binning, and other properties as needed
  5. Add additional attributes to the legend to encode with color, size, or shape
  6. Use the ruler tool to create groupings or reference lines
  7. Add text objects with titles, labels, or explanatory notes
  8. Arrange multiple visualizations to create a dashboard

Activity 27. Creating Visualizations in CODAP.

In this activity, you’ll practice creating various types of visualizations in CODAP using your dataset.
(a)
Create at least one of each of the following visualization types using your dataset:
  • Scatter plot showing the relationship between two numerical variables
  • Bar chart or dot plot comparing values across categories
  • Histogram showing the distribution of a numerical variable
  • Box plot comparing distributions across groups
(b)
For each visualization:
  • Add appropriate titles and labels
  • Adjust scales and binning as needed
  • Use color or other visual encodings to add an additional dimension
  • Write a brief interpretation of what the visualization reveals
(c)
Arrange your visualizations to create a simple dashboard with at least three different charts that together tell a story about your data.

Checkpoint 93. CODAP Visualization Features.

    Which CODAP feature would be MOST appropriate for examining how a relationship between two numerical variables differs across categories?
  • Changing from a dot plot to a histogram
  • Changing to a histogram would show the distribution of a single variable, not the relationship between two variables across categories.
  • Adding connecting lines between points
  • Connecting lines are typically used to show sequences or paths between related points, not to distinguish between categories.
  • Dragging a categorical attribute to the legend area to color points by category
  • Correct! By dragging a categorical attribute to the legend area, points in the scatter plot will be colored according to their category, allowing you to visually distinguish how the relationship differs across categories.
  • Using the ruler tool to add reference lines
  • While reference lines can be helpful markers, they don’t specifically help examine how relationships differ across categories.
PrevTopNext