Introduction to Critically Assessing Visualisations
In our data-driven world, we are surrounded by graphs, charts, and infographics. While these are designed to make data easier to understand, they can often be misleading—sometimes by accident, and sometimes on purpose! As an A Level Statistics student, your job isn't just to read a graph, but to be a "stat-detective." In this chapter, we will learn how to spot the "dirty tricks" used in published visualisations and how to justify why one type of graph might be better than another.
This chapter is part of the Data Analysis and Presentation section. It focuses on AO3 (Assessment Objective 3): critically assessing the reliability and validity of how data is shown.
What Makes a "Good" Visualisation?
Before we can spot a bad graph, we need to know what a good one looks like. According to the Pearson Edexcel syllabus, an appropriate representation should be clear, honest, and easy to interpret. Key features include:
- Clear Titles and Labels: Does the graph tell you exactly what it is showing, including units (e.g., cm, kg, \(\text{£}\))?
- Appropriate Scales: Are the increments on the axes consistent?
- A Key or Legend: If there are multiple data sets (multivariate data), is it clear which is which?
- The Source: Do we know where the data came from? Is the source biased?
Common Ways Visualisations Mislead
When you are asked to critically assess a published visualisation in your exam, look for these common "traps":
1. The Truncated Axis (The "Broken" Y-Axis)
This is when the vertical axis does not start at zero. This trick is often used to make small differences look huge.
Example: A bar chart showing a rise in house prices from \(\text{£}200,000\) to \(\text{£}202,000\). If the axis starts at \(\text{£}200,000\), the second bar will look twice as tall as the first, even though the increase is only \(1\%\)!
2. Unequal Class Widths in Histograms
In a histogram, the area of the bar represents the frequency, not the height. A common mistake is plotting frequency on the y-axis instead of frequency density when class widths are different.
Remember: \( \text{Frequency Density} = \frac{\text{Frequency}}{\text{Class Width}} \). If a graph uses frequency on the y-axis for unequal classes, it is misleading!
3. Improper Scaling of Icons (Pictograms)
If a pictogram uses images to represent data, the images must be scaled by area, not just height. If you double the height and width of a square icon, the area becomes \(4\) times larger (\(2^2\)), making the increase look much bigger than it actually is.
4. Cherry-Picking Timeframes
On a time series graph, a presenter might only show a small window of time to hide a larger trend. For example, showing that profits rose over three months, while hiding the fact that they have been falling for three years.
Assessing Multivariate Visualisations
The syllabus specifically mentions multivariate visualisations. "Multivariate" simply means the graph shows more than two variables at once. For example:
- Scatter Diagrams with Regression Lines: These show the relationship between two variables, but might use different colors or symbols for a third variable (e.g., gender or age group).
- Comparative Box Plots: Showing multiple box and whisker plots on the same scale to compare different populations.
- Time Series with Multiple Lines: Showing how several different categories change over the same time period.
Critical Tip: When assessing these, check if the comparison is "fair." Are they using the same scale? Is the legend clear?
The Statistical Enquiry Cycle (SEC) Connection
Critically assessing data falls under the "Interpretation" and "Evaluation and Review" stages of the Statistical Enquiry Cycle. When you look at a published graph, ask yourself these SEC-inspired questions:
1. Bias: Was the data collected in a way that avoids bias? (Check back to the Population and Samples chapter for more on this!)
2. Suitability: Is a pie chart the best choice for this data? (Pie charts are often poor for comparing small differences; a bar chart might be better).
3. Reliability: Is the sample size (\(n\)) large enough to justify the conclusions shown in the graph?
Quick Review: The "Checklist" for Exam Questions
If you are asked to comment on or critically assess a visualisation, try to find at least one positive and one negative point. Use this checklist:
Check the Axes: Do they start at \(0\)? Are the scales linear (going up in equal steps)?
Check the Labels: Are there units? Is there a title?
Check the Type: Is it a histogram? If so, check the y-axis label for "Frequency Density."
Check the Data: Are there outliers that have been ignored? (Recall: \( \text{Outlier} > \text{UQ} + 1.5 \times \text{IQR} \)).
Check the Context: Does the graph actually answer the question it claims to answer?
Don't worry if this seems tricky at first! Identifying misleading graphs is a skill that gets easier with practice. In the exam, use common sense—if a graph looks like it's trying to "scare" you or "sell" you something, look closer at the scales!
Key Takeaways
- Visualisations must be clear, labeled, and use appropriate scales to be valid.
- Misrepresentation often occurs through truncated axes, incorrect area scaling, or failing to use frequency density in histograms.
- Multivariate data requires careful comparison using keys and consistent scales.
- Critical assessment is part of the "Evaluation" stage of the Statistical Enquiry Cycle (SEC).
- Always interpret in context: Mention the actual variables (e.g., "The y-axis for 'Temperature' starts at \(15^{\circ}C\), which exaggerates the heat increase").