Statistics and Data Visualisation
Core descriptive measures, variation, relationships, and chart selection
These original notes are designed for study and skill development. They are not certificates, official records, or
copied course materials.
1. Describing a dataset
Descriptive statistics summarise the centre, spread, and shape of data. No single measure tells the complete story,
so centre and variation should usually be reported together.
Measure Meaning Best use
Mean Arithmetic average Balanced numeric data without severe outliers
Median Middle ordered value Skewed data such as income or delivery time
Mode Most frequent value Categories or repeated discrete values
Range Maximum minus minimum Quick spread check, sensitive to extremes
Interquartile range Middle 50% spread Robust comparison of skewed distributions
Standard deviation Typical distance from the mean Variation in roughly symmetric numeric data
Distribution shape
- A symmetric distribution has similar shape on both sides of its centre.
- A right-skewed distribution has a longer high-value tail.
- Outliers may be errors, rare valid events, or important signals.
- Multiple peaks can indicate hidden subgroups.
Example: A median delivery time of 24 minutes and a 90th percentile of 61 minutes shows that most deliveries are
faster, but a meaningful minority take much longer.
Statistics and Data Visualisation Page 1
2. Relationships and uncertainty
Correlation
Correlation measures the strength and direction of association between two variables. A value near +1 indicates a
strong positive linear relationship, near -1 a strong negative linear relationship, and near 0 little linear relationship.
- Correlation can miss curved relationships.
- A strong correlation does not prove causation.
- Outliers can change the coefficient substantially.
- Aggregated data can hide different subgroup patterns.
Sampling and confidence
A sample estimate varies from sample to sample. A confidence interval describes a range produced by a method
that would capture the true population value at a stated long-run rate under its assumptions.
Hypothesis-testing idea
Term Meaning
Null hypothesis A baseline claim such as no difference or no effect
Alternative hypothesis The competing claim supported when evidence is strong enough
p-value How unusual the observed result would be under the null model
Significance level A preselected evidence threshold, often 0.05
Effect size The practical magnitude of the difference or relationship
Do not report only a p-value. Include the estimated effect, uncertainty, sample size, assumptions, and practical
importance.
Statistics and Data Visualisation Page 2
3. Choosing effective charts
Analytical goal Good starting chart Avoid
Compare categories Sorted bar chart 3D bars and excessive categories
Show trend over time Line chart Unordered dates
Show distribution Histogram or box plot Only reporting the average
Show relationship Scatter plot Connecting unrelated points
Show composition Stacked bar or 100% stacked bar Many-slice pie charts
Show geographic pattern Map when location is central Maps for simple ranking
Design principles
- Use a title that states the insight or exact question.
- Label units and time periods clearly.
- Start bar-chart axes at zero unless a different choice is clearly justified.
- Use colour to highlight meaning, not decoration.
- Remove unnecessary gridlines, borders, and repeated legends.
- Show source, filters, and important caveats.
Practice
- Choose a chart for monthly revenue across two years.
- Choose a chart for customer waiting-time distribution.
- Explain why a pie chart with 18 products is difficult to read.
- Describe one misleading axis choice.
- Write an insight-focused title for a chart showing a 14% fall in returns.
Communication test: A reader should be able to identify the main finding, comparison, scale, and limitation without
asking how to decode the chart.
Statistics and Data Visualisation Page 3