Judging the quality of a dataset

Name:

Date:

Class:

Quick Questions

Shade in the A or B circle for each question.

1. A student wants to judge the quality of a dataset. Which approach matches a proper data quality check?

A Test accuracy, bias, integrity, relevance and reliability one at a time

B Glance at the dataset and decide if it looks good overall

2. A dataset of student heights lists one Year 10 student as 17 m tall. Which quality characteristic is this evidence against?

A Accuracy

B Relevance

3. A survey about school canteen food was only given to students who were already in the canteen queue. What is the best quality judgement?

A No problem, because every response was recorded correctly

B Possible bias, because the sample leaves out students who do not buy canteen food

4. Which statement is the stronger way to explain a quality problem?

A This dataset is bad and should not be trusted

B Row 14 has no postcode, so the data is incomplete and suburb totals could be too low

5. A dataset about bus delays in Melbourne is used to answer a question about train delays in Perth. Which characteristic is the main problem?

A Relevance

B Integrity

Match the Pairs

Draw a line connecting each item on the left to its match on the right.

Left column		Right column
1Accuracy check: 3 of 40 ages are 250		ABias problem: the sample over-represents one group, so conclusions about the whole school may be unfair
2Bias check: 90% of replies are from Year 12		BJudging quality by appearance only. Each characteristic must be tested separately using evidence, such as counting errors.
3Integrity is checked by looking for missing, duplicated or mismatched records, and giving evidence by counting them		CAccuracy problem: some values are impossible or wrong, so averages and totals drawn from them could be misleading
4Relevance is checked by asking whether each field helps answer the question, and noting data that does not		DIntegrity check: 12 blank cells, 5 duplicate IDs
5Quality check: The data looks neat, so it is high quality.		EReliability needs evidence of consistent results, such as repeated surveys or a trusted source. One result proves nothing.
6Reliability check: One survey gave this result, so it is reliable.		FRelevance check: sleep data used to study canteen sales

Definitions

data quality check

Fill in the Blanks

Complete each line of code by filling in the missing parts.

=_______(B2:B100,">120")

=__________(C2:C100)

=COUNTIF(D2:D100,"Year 12")/______(D2:D100)

=__(B2=C2,"Match","Check")

=______(B2:B100)

Spot the Error

Shade in the letter of the line that contains an error.

Question 1

A dataset ← survey_results

B accuracy ← no rows missing

C evidence ← compared 20 rows

Question 2

A check ← bias

B ask: were values typed right?

C evidence ← only Year 12 asked

Question 3

A check ← relevance

B purpose ← plan canteen menu

C IF data is large: keep

D ELSE: discard it

Question 4

A check ← reliability

B repeat ← collect data again

C IF results are similar

D reliable ← FALSE

E ELSE reliable ← FALSE

Question 5

A check ← integrity

B missing ← count blank cells

C IF missing = 0: flag problem

D ELSE: no integrity problem

Answer Key

Judging the quality of a dataset

Section A: Quick Questions

A — Test accuracy, bias, integrity, relevance and reliability one at a time

A — Accuracy

B — Possible bias, because the sample leaves out students who do not buy canteen food

B — Row 14 has no postcode, so the data is incomplete and suburb totals could be too low

A — Relevance

Section B: Match the Pairs

Accuracy check: 3 of 40 ages are 250 → Accuracy problem: some values are impossible or wrong, so averages and totals drawn from them could be misleading

Bias check: 90% of replies are from Year 12 → Bias problem: the sample over-represents one group, so conclusions about the whole school may be unfair

Integrity is checked by looking for missing, duplicated or mismatched records, and giving evidence by counting them → Integrity check: 12 blank cells, 5 duplicate IDs

Relevance is checked by asking whether each field helps answer the question, and noting data that does not → Relevance check: sleep data used to study canteen sales

Quality check: The data looks neat, so it is high quality. → Judging quality by appearance only. Each characteristic must be tested separately using evidence, such as counting errors.

Reliability check: One survey gave this result, so it is reliable. → Reliability needs evidence of consistent results, such as repeated surveys or a trusted source. One result proves nothing.

Section C: Definitions

data quality check — Judging a dataset by testing accuracy, bias, integrity, relevance and reliability one at a time, and using evidence to show how any problem could affect conclusions.

Section D: Fill in the Blanks

=COUNTIF(B2:B100,">120")

=COUNTBLANK(C2:C100)

=COUNTIF(D2:D100,"Year 12")/COUNTA(D2:D100)

=IF(B2=C2,"Match","Check")

=MEDIAN(B2:B100)

Section E: Spot the Error

B — Accuracy is about whether values are correct, so the check is that values match the source. Checking that no rows are missing tests integrity (completeness), not accuracy. The line should be: accuracy ← values match source.

B — Checking whether values were typed correctly tests accuracy. A bias check asks who or what was left out or over-represented. The line should be: ask: who was left out?

C — Relevance means the data suits the purpose of the task, not how large it is. A large dataset can still be irrelevant. The line should be: IF data fits purpose: keep.

D — If repeating the collection gives similar results, the data is reliable. Setting reliable to FALSE contradicts the IF condition. The line should be: reliable ← TRUE.

C — A problem should be flagged when blank cells exist (missing is more than 0), not when there are none. The line should be: IF missing > 0: flag problem.