Reference

Core Terms

Term	Meaning in this course
Observation	The entity or event represented by one row, such as one request or one measurement at a stated time
Variable	A property recorded for each observation, with its meaning, unit, and allowed values defined
Identifier	A value used to distinguish or link records; check whether it is expected to be unique
Data dictionary	Documentation of column meanings, types, units, categories, and missing-value conventions
Missing value	An absent observation; its reason may matter, and it is not automatically zero
Validation	Checks against stated rules, source records, or expected relationships
Duplicate candidate	Records that may repeat the same observation; matching values alone do not establish duplication
Provenance	The source of the data and the record of how it reached its current form
Selection bias	A difference caused by who or what is included in the data, affecting what it can represent
Exploratory analysis	Inspecting structure, distributions, and relationships to understand data and develop questions
Denominator	The population or count used to calculate a proportion, rate, or average
Outlier	An observation unusual relative to others; investigate its context before treating it as an error
Association	Variables vary together; this alone does not establish that one causes the other
Uncertainty	Limits to what a result establishes, including sampling variation, measurement issues, or missing context

Collecting the Right Data

Use with Collecting the Right Data.

Before collecting or reusing data, record:

1. The question and decision the data should inform.
2. What one row will represent, including any repeated measurements.
3. The population of interest and who or what the source may miss.
4. The variables needed, their units, categories, and collection method.
5. The source, collection dates, permitted use, and important limitations.

Do not equate a large dataset with a representative dataset. Consider whether a source covers the people, places, period, or events in your question.

Organising and Exploring Data

Use with Organising and Exploring Data.

Before changing values

• Keep the original file and work on a copy.
• Preserve stable row or observation identifiers.
• Define missing values explicitly and keep them distinct from zero.
• Check units and categories against documentation before converting them.
• Investigate duplicate candidates against the collection process and source records.
• Do not replace an unusual value with a plausible guess.

Use a change log with record identifier; original value; change; evidence; reason. Keep unresolved issues visible. A tidy layout does not establish that the measurements are valid.

Before trusting a summary

Check	Why it matters
Row count and unique observation count	Detects unplanned row loss, repeated exports, or legitimate repeated measurements
Missingness by column and group	Shows which observations contribute to a comparison
Minimum, maximum, and units	Helps identify impossible values or mixed scales
Category counts	Reveals spelling variants and unexpected categories
Mean, median, and distribution	Shows how an average may conceal skew, spread, or unusual observations
Filters and denominator	Makes exclusions and the basis of a result explicit

Mean: sum of the included values divided by their count. Median: the middle value after sorting; average the two middle values for an even count. Mode: the most frequent value or category; ties are possible. State which records contribute to each calculation.

Making Sense of Data

Use with Making Sense of Data.

Question	Useful starting point	Check
How do categories compare?	Bar chart	Show counts or rates clearly; bar length should use a meaningful baseline
How does a measure change over time?	Line chart	Label time intervals and account for gaps or changes in measurement
What is the distribution of a numerical measure?	Histogram or box plot	Check binning or plot conventions; retain information about spread
Are two numerical measures associated?	Scatter plot	Consider unusual observations, group differences, and alternative explanations

Check the title, source, units, period, denominator, filters, uncertainty, and axis scale. Add a text description or data table so the evidence can be understood without relying on the image or colour alone. See the Government Analysis Function chart guidance for worked design and accessibility examples.

Insight to Impact

Use with Insight to Impact.

Keep four statements distinct:

Statement	What to record
Evidence	What was observed, in which source and population, over which period
Interpretation	What the observation may mean, with alternative explanations
Proposed action	Who could do what, with constraints and possible unintended effects
Evaluation	How you will check whether the action helped, for whom, and under what conditions

A before-and-after change can support further investigation, but other changes may explain it. Do not label an action effective solely because a result improved after it was introduced.

End-of-Course Checklist

You should be able to:

• connect a collection plan to a clear question and population
• define row meaning, identifiers, variables, units, and missing values
• explain and document justified cleaning decisions
• calculate a summary with a stated denominator and exclusions
• choose and critique a chart using its source and context
• distinguish an observation, an interpretation, and a proposed action
• plan feedback that checks the effects and limitations of an action

Practise these checks with the worked fictional dataset.

Related Foundations

For storage, metadata, ethics, and institutional research systems, use the Essential Digital Skills reference. Those foundations support this course; its four lessons focus on data collection, exploration, interpretation, and impact.

For a coding route after this course, see Applied Python or Applied R.

Course Navigation

Course Overview · Setup · Additional Material · Reference · Discussion