Data Visualization Guide: Chart Types and Data Types for Students
At the 2024 AI Connect K-12 Data Science Conference, KPMG's Alexander Baker explained how to choose among histograms, bar, pie, scatter, bubble, time and geospatial charts. He tied each choice to the data type and warned that too many categories and outliers hurt readability.
By The AI Enterprise Magazine Team3 min read
Key takeaways
- Know your data first. Baker split data into nominal, ordinal, interval and ratio, and discrete versus continuous.
- Match chart to data. Histograms suit one numerical variable, bar and pie charts one categorical variable, scatter plots two numerical ones.
- Watch for scale problems. Two features on very different scales can make one look flat on a line chart.
- Readability wins. Too many categories or outliers can make a chart hard to read.
- Tell a story. Baker said charts are building blocks of a narrative that can be assembled on a dashboard.
In a 2024 AI Connect K-12 Data Science Conference session, Alexander Baker, Senior Federal Data Scientist at KPMG, walked students through the basics of data visualization: the types of data they will meet, the chart that suits each one, and the pitfalls that make a graphic hard to read. The virtual talk was aimed at students new to data work.
Baker defined data visualization as "the practice of representing data through some sort of visual medium." Charts, he said, are the building blocks of a story, and they can be assembled on a dashboard.
Why does data visualization matter?
Baker said people process visuals far faster than text, citing a figure of about 60,000 times faster. He did not give a source for the number in the talk. The speed, he said, lets people gain insight quickly and makes visualization a strong foundation for informed decision making.
What types of data should you know before making a chart?
Before choosing a chart, Baker said, it helps to know what kind of data you have. He split data into two families.
Categorical data consists of labels and names. It comes in two kinds:
- Nominal, which has no order, such as student names or school subjects.
- Ordinal, which is labeled but ordered, such as letter grades, where an A is better than a B.
Numerical data consists of numbers, such as height or weight. Baker described two further splits:
- Interval versus ratio. A ratio scale has a true zero, meaning zero stands for the absence of anything. He said Fahrenheit is an interval scale, while Kelvin is a ratio scale because it has an absolute zero.
- Discrete versus continuous. Discrete data is counted, like marbles in a bag. Continuous data is measured, like weight or height.
Which chart should you use for which kind of data?
Baker said the first step is to keep your goal in mind, then pick the technique that fits the data. He covered eight chart types:
- Histogram: shows the distribution of one numerical variable by grouping values into buckets. Outliers or dominant buckets can skew the picture.
- Bar chart: shows the frequency of values in a single categorical variable. Too many categories hurt readability, and one large category can overshadow the others.
- Pie chart: similar to a bar chart but used to show the prevalence of categories. Baker said it has the same pitfalls.
- Stacked bar chart: shows the relationship between two categorical variables. His example was animals at a pound: cats and dogs as two bars, each split into adopted and not adopted.
- Scatter plot: shows the relationship between two numerical features on an X and Y plane.
- Bubble chart: a scatter plot where dot size adds a third numerical feature.
- Time chart: plots one numerical feature over a period, often as a line. Baker warned that when two features sit on very different scales, one can understate the behavior of the other.
- Geospatial plot: maps a feature by latitude and longitude, such as poverty or energy consumption across countries or counties. Very small land areas can be hard for viewers to see.
What are the common mistakes in data visualization?
Across all chart types, Baker said, readability comes first. Keep the number of categories, points and features at a manageable level so the viewer can understand the point. Outliers can bury information and make a plot difficult to read, he said. He also advised keeping titles, legends and text concise and informative.
His closing advice was to "always remember to focus on the readability and your overall narrative" when building charts. Students will meet these concepts in labs, classes and offices, he said, since they apply wherever information is involved. He pointed to further examples of advanced charts and software for building charts quickly as next steps.
Learn more with AI Connect
Data visualization is a theme across AI Connect programming. The organization's 24 free live webinars on the 2027 schedule run in four series: the AI × Enterprise Panel Series, AI Demo Days, Agentic AI Showcases and the Plug & Play Series. See the webinars page for details. Registration for 2027 events is coming soon.
Frequently asked questions
What is data visualization?
Alexander Baker defined it as representing data through a visual medium, with charts serving as building blocks for a story or dashboard.
What is the difference between nominal and ordinal data?
Both are categorical. Nominal data, like student names, has no order. Ordinal data, like letter grades, is labeled and ordered.
When should you use a scatter plot?
Baker said scatter plots show the relationship between two numerical features, and bubble charts add a third through dot size.
What are common data visualization mistakes?
Baker named too many categories, dominant categories or outliers, mismatched scales and wordy titles or legends.
This article is based on remarks by Alexander Baker, Senior Federal Data Scientist, KPMG at AI Connect K-12 Data Science Conference 2024 on July 25, 2024. More coverage: More AI Connect Events.
Filed under: AI Use Cases & Applications, People, Culture & Leadership
