eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

Why English? | Reading a Data Dictionary Before Joining Two Datasets

—

by

eduKate Asia
DATA DICTIONARY · JOIN KEYS · PROVENANCE

Make two datasets describe the same thing before you connect them

Name both datasets, open metadata, test identifiers and cardinality, align time and units, preserve unmatched rows and publish the join as your own traceable transformation.

ROUTE 1IdentifyName datasets, periods and official owners.ROUTE 2MatchRead fields, keys, types and cardinality.ROUTE 3JoinChoose the operation from the question.ROUTE 4AuditValidate rows, unmatched keys and provenance.

Open the full chapter index ↓

Why this English matters now

Reading a data dictionary before joining two datasets is a practical English, data literacy, research and workplace skill.

People searching data dictionary, metadata, join datasets, merge CSV files or join key need to know what each column means before asking software to connect rows.

A successful command can still create a false comparison when identifiers, time periods, units, populations or geographic levels differ. Strong readers test meaning and cardinality before calculation.

English matters because definitions, footnotes and state labels determine the relationship. This guide uses current Singapore government data tools as examples and does not make either agency responsible for a user's derived analysis.

Checked against the linked first-party sources on 29 October 2026. This educational guide explains how to read and write more precisely; it does not replace the organisation, programme owner, technology provider or government agency responsible for a live decision.

The dictionary-to-join route

Use this sequence before a live decision.

  1. Record dataset identity, coverage and intended question.
  2. Open metadata and define every candidate join field.
  3. Test uniqueness, cardinality, units, types and boundaries.
  4. Choose a join type and inspect unmatched or multiplied rows.
  5. Validate samples and publish transformation, provenance and limits.

Official factual floor

These primary sources establish the current factual examples and should be reopened before a live application, update, citation or submission:

  • data.gov.sg distinguishes collections from datasets and provides metadata and download APIs.
  • SingStat says definitions and footnotes are available in the table view or metadata panel and can be included with downloads.
  • SingStat's developer API separates resource discovery, metadata and table-data endpoints.
Scope: This guide does not certify a join, define every publisher's schema, replace statistical advice or authorise personal-data processing. Reopen the live metadata and applicable licence before using data.

Did you know?

A join can run without an error while multiplying the wrong rows. The most important error message may therefore be the row count you calculate yourself.

Full chapter index

  1. 1. Name both datasets before joining them
  2. 2. Open the metadata and data dictionary
  3. 3. Find a real key rather than a convenient label
  4. 4. Check uniqueness on both sides
  5. 5. Align time, geography and population
  6. 6. Translate units and data types deliberately
  7. 7. Read null, zero and suppressed values correctly
  8. 8. Normalise labels without erasing meaning
  9. 9. Choose the join type from the question
  10. 10. Audit row counts and unmatched keys
  11. 11. Validate a sample against the source
  12. 12. Publish the join with provenance and limits
CHAPTER 1 OF 12

1. Name both datasets before joining them

Back to contents

How the dictionary controls the join

Write down each official title, dataset identifier, publisher and applicable period. A join is only meaningful after both inputs are unambiguous.

Field names are clues, not definitions. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A student joins a current resident-population file to an older planning-area table because both contain a column called area.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. copy both titles
  2. 2. record both identifiers
  3. 3. record coverage dates
  4. 4. state the intended relationship

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Similar filenames or column names do not prove that two datasets describe the same population, time or geography.

Dataset identity supports reproducible science, school investigations and workplace analytics.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 2 OF 12

2. Open the metadata and data dictionary

Back to contents

How the dictionary controls the join

Use the publisher's metadata, definitions, footnotes or API schema to learn what each field means. On SingStat, definitions and footnotes can be viewed in the metadata panel and included with downloads.

A key needs level, uniqueness and time. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A spreadsheet column labelled value is treated as dollars although its metadata defines a number of persons.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. open metadata
  2. 2. find every join field
  3. 3. copy unit and definition
  4. 4. keep footnotes beside the analysis

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: A familiar English word can carry a specialised statistical definition.

Metadata reading transfers to laboratory tables, surveys and administrative records.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 3 OF 12

3. Find a real key rather than a convenient label

Back to contents

How the dictionary controls the join

A join key should identify the intended entity at the required level. Stable codes usually work better than display names because punctuation, spelling and abbreviations can vary.

Keep raw and cleaned values together. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: Bukit Merah and BUKIT MERAH appear to match, but one file uses planning areas and the other uses subzones.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. identify candidate key
  2. 2. check uniqueness
  3. 3. check level
  4. 4. prefer stable official codes

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: A readable label is not automatically a unique identifier.

Key selection supports databases, contact lists and course records.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 4 OF 12

4. Check uniqueness on both sides

Back to contents

How the dictionary controls the join

Test whether the key appears once, many times or not at all in each dataset. One-to-one, one-to-many and many-to-many joins produce different row counts and meanings.

Audit what did not match. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A school list has one row per school, while a programme file has many rows per school; joining without recognising this multiplies records.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. count key frequencies
  2. 2. name join cardinality
  3. 3. inspect duplicates
  4. 4. decide whether aggregation is needed

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Duplicate keys can be valid observations rather than errors.

Cardinality reasoning prevents inflated totals and duplicated evidence.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 5 OF 12

5. Align time, geography and population

Back to contents

How the dictionary controls the join

A shared code does not guarantee shared coverage. Compare reference period, release status, geographic boundary and population definition before joining.

Field names are clues, not definitions. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A 2025 household dataset is joined to a 2020 boundary file and described as a current neighbourhood comparison.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. align period
  2. 2. align geography
  3. 3. align population
  4. 4. record any unavoidable mismatch

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: A technically successful join can still be analytically false.

Boundary alignment supports responsible education and career research.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 6 OF 12

6. Translate units and data types deliberately

Back to contents

How the dictionary controls the join

Check whether a field is text, integer, date, rate, percentage or currency, and whether leading zeroes or missing values carry meaning.

A key needs level, uniqueness and time. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A postal-sector code loses a leading zero when a spreadsheet converts it to a number.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. inspect field type
  2. 2. preserve identifiers as text
  3. 3. standardise date format
  4. 4. convert units explicitly

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Changing type for convenience can change identity or precision.

Type discipline supports forms, coding and digital administration.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 7 OF 12

7. Read null, zero and suppressed values correctly

Back to contents

How the dictionary controls the join

Metadata may distinguish missing, not applicable, negligible, provisional and true zero values. Keep those states separate before matching or calculating.

Keep raw and cleaned values together. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A dash meaning nil or negligible is converted to blank and later treated as missing data.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. list special symbols
  2. 2. map each state
  3. 3. preserve raw value
  4. 4. document any recoding

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Blank, zero and unavailable are not interchangeable.

State-aware reading supports surveys, assessments and dashboards.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 8 OF 12

8. Normalise labels without erasing meaning

Back to contents

How the dictionary controls the join

Cleaning can trim spaces, standardise case and map documented aliases, but every transformation should be visible and reversible.

Audit what did not match. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: Two course titles are forced to match after punctuation removal even though one includes a different award level.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. keep raw fields
  2. 2. create cleaned fields
  3. 3. use documented mappings
  4. 4. review changed matches

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Fuzzy matching suggests candidates; it does not certify identity.

Transparent cleaning supports qualification and career comparisons.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 9 OF 12

9. Choose the join type from the question

Back to contents

How the dictionary controls the join

An inner join keeps matches only; left, right and full joins preserve different unmatched records. Select the operation because of the research question, not because it is the software default.

Field names are clues, not definitions. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: An inner join silently removes schools that have no record in the second file, making coverage look complete.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. state desired population
  2. 2. choose join type
  3. 3. count unmatched rows
  4. 4. explain exclusions

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Removing unmatched records can create selection bias.

Join-type literacy supports statistics and evidence-based decisions.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 10 OF 12

10. Audit row counts and unmatched keys

Back to contents

How the dictionary controls the join

Compare counts before and after the join, inspect duplicated rows and list unmatched keys from both sides. Unexpected changes are clues that the key, coverage or cardinality needs attention.

A key needs level, uniqueness and time. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A 500-row source becomes 1,400 rows, yet the team calculates a total without investigating.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. record input counts
  2. 2. record output count
  3. 3. inspect unmatched keys
  4. 4. explain multiplication

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: A matching percentage alone cannot prove that matched rows are correct.

Audit trails support peer review and dependable reporting.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 11 OF 12

11. Validate a sample against the source

Back to contents

How the dictionary controls the join

Pick known entities, trace their original rows and confirm that the joined values describe the same entity, period and unit.

Keep raw and cleaned values together. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: The join passes automated checks, but a manual sample reveals that school codes were recycled across historical files.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. choose representative cases
  2. 2. trace both originals
  3. 3. check joined sentence
  4. 4. record exceptions

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: Manual validation is a sample, not proof that every row is correct.

Evidence tracing supports research, applications and operational data.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

CHAPTER 12 OF 12

12. Publish the join with provenance and limits

Back to contents

How the dictionary controls the join

Name the two datasets, identifiers, access dates, dictionary fields, cleaning rules, join type, unmatched records and major limitations.

Audit what did not match. This extra pass matters because a short label often compresses several decisions: who the information applies to, which version is current, what condition must be met and what action follows. English becomes useful when those relationships remain visible instead of being flattened into a vague impression.

A dataset-joining example

Example: A polished chart gives no indication that 12 percent of records failed to match.

The mistake is understandable, but it changes the decision. A careful reader returns to the authoritative page, copies the exact governing words, checks the date and scope, and writes a short note that separates what the source says from what the reader infers. That note makes later discussion calmer because everyone can inspect the same evidence.

Four careful reading moves
  1. 1. cite both sources
  2. 2. describe key and join
  3. 3. report unmatched records
  4. 4. state limitations

After the four moves, explain the result in one complete sentence. Include the document or page, the relevant state, the condition and the next action. If the sentence sounds more certain than the source, narrow it. If it is too vague to guide a responsible action, return to the source and locate the missing qualifier.

What a successful join cannot prove

Important boundary: A joined dataset is a derived product; do not imply that either agency published or endorsed it.

Provenance makes data work usable by classmates, parents and colleagues.

For a student, the transfer may appear in an assignment, application or group project. For a parent, it may shape a deadline or consent decision. For a teacher, adviser or manager, it may determine whether a question can be answered from the record or must be routed to the issuing body. The shared habit is simple: preserve source, version, scope and action together.

Four discoveries about data joins

Names are not keys

Display labels can vary or repeat.

Use it: Prefer documented identifiers.

Cardinality predicts row count

One-to-many and many-to-many joins multiply records.

Use it: Test frequency first.

Unmatched rows are evidence

They reveal coverage, spelling or boundary differences.

Use it: Report them.

A join creates a new product

The combined table reflects your decisions.

Use it: Publish provenance.

Dataset-join practice lab

Dictionary walk

Define five fields from official metadata.

  • name
  • type
  • unit
  • state

Cardinality test

Classify one-to-one, one-to-many and many-to-many examples.

  • left key
  • right key
  • duplicates
  • expected rows

Join audit

Explain every unmatched or multiplied row class.

  • count
  • cause
  • action
  • limit

Advice for students, families and educators

For students

  • Keep metadata beside the downloaded file.
  • Report unmatched rows.
  • Do not hide a boundary mismatch.

For families

  • Ask what the shared field actually identifies.
  • Encourage a modest conclusion.

For educators and analysts

  • Teach row-count audits before charts.
  • Require source and transformation notes.

Frequently asked questions

What is a data dictionary?

A source that defines fields, codes, units and states in a dataset.

Is a column name enough?

No; read its official definition and footnotes.

What is a join key?

A field or fields used to match the intended entities.

Why check uniqueness?

It reveals whether rows can multiply.

What is an inner join?

It keeps only matching records from both inputs.

Are blank and zero the same?

No.

Can I clean names?

Yes, transparently, while preserving raw values.

Why inspect unmatched rows?

They reveal coverage and mapping problems.

Who owns the joined result?

Describe it as your derived analysis and follow applicable licences.

Why English?

Because definitions and footnotes decide what the data can mean together.

Useful next reading

Closing thought

The safest join begins as a reading exercise: define each field, prove the relationship, keep the records that resist matching and let the published analysis show exactly how two sources became one table.