4.1 Sampling, data quality and outliers
Studying this course? Keep your progress in one place.
Key facts
- A population is everyone studied; a sample is the part used to draw conclusions.
- Discrete data are counted; continuous data are measured.
- Bias comes from how members are chosen, not from the sample being small.
- Stratified sampling: group size population sample size.
- An outlier lies more than below or above .
Population, sample and data
A population is the whole group; a sample is a part of it, and data are discrete if counted or continuous if measured.
A population is the whole group being studied. A sample is a smaller part used to draw conclusions about the whole. A random sample gives every member an equal chance of selection. At SL, a data set in a question is the population unless stated.
Discrete data take separate values, usually counted. Continuous data take any value in a range, usually measured.
Cars passing a school in an hour are discrete; the time a car takes to pass a point is continuous.
- 1
Population
The whole group being studied.
- 2
Take a sample
A random sample gives every member an equal chance.
- 3
Collect data
Discrete data are counted; continuous data are measured.
- 4
Draw conclusions
Use the sample to say something about the population.
| Discrete | Continuous | |
|---|---|---|
| Values | Separate | Any value in a range |
| How collected | Counted | Measured |
| Examples | Siblings, goals | Height, time, mass |
Discrete
- Values:
- Separate
- How collected:
- Counted
- Examples:
- Siblings, goals
Continuous
- Values:
- Any value in a range
- How collected:
- Measured
- Examples:
- Height, time, mass
The mass of a parcel is which type of data?
Reliability and bias
A source is reliable if trustworthy; bias comes from how members are chosen, so a large sample cannot cure it.
A source is reliable if it comes from a reputable organisation, uses a clear method and a large enough sample, and gives similar results when repeated.
Bias is a systematic tendency for a sample to differ from the population. Causes include members with no chance of being chosen, leading questions and only some people replying (non-response).
A larger sample does not remove bias: a large sample from the wrong group is still biased.
Unreliable
- Small sample
- Unclear method
- Results change on repeat
Biased
- Some members cannot be chosen
- Leading questions
- Non-response
A survey of 5000 people is taken only outside a gym. Its main problem is...
Random, systematic and stratified
Random sampling gives equal chances, systematic takes every th member, and stratified samples each group in proportion.
Simple random sampling: number every member and use random numbers (GDC) to pick distinct members. Fair, but needs a full list and may miss small groups.
Systematic sampling: take every th member from a list, starting at a random place, with . Quick, but biased if the list has a pattern matching .
Stratified sampling: split the population into groups (strata) and sample each in proportion to its size.
- Sample from a group
Worked example
A club has 600 members: 240 juniors, 210 adults and 150 seniors. Choose a stratified sample of 80.
- 1
Each group gets .
- 2
Juniors: . Adults: .
- 3
Seniors: .
- 4
Check the total: .
A population of 1000 has a sample of 50 taken systematically. What is ?
Quota and convenience sampling
Quota copies the group structure but picks people non-randomly; convenience takes whoever is easiest to reach.
Convenience sampling takes whoever is easiest to reach, such as the first people to walk past. It is cheap and fast but often biased.
Quota sampling has the interviewer find a set number in each group (say 20 men and 20 women) but choose them non-randomly. It reflects the group structure like stratified sampling, but no sampling frame is needed and bias is still possible.
Random and stratified samples are usually the most representative; convenience samples the least.
- 1
Decide the groups and quotas
For example 20 men and 20 women.
- 2
Interviewer chooses people
Not randomly: whoever is available.
- 3
Stop when each quota is full
The sample copies the group structure.
- 4
Bias is still possible
No sampling frame is used.
| Stratified | Quota | Convenience | |
|---|---|---|---|
| Uses groups | Yes | Yes | No |
| Random selection | Yes | No | No |
| Main weakness | Needs group sizes | Bias within groups | Often unrepresentative |
Stratified
- Uses groups:
- Yes
- Random selection:
- Yes
- Main weakness:
- Needs group sizes
Quota
- Uses groups:
- Yes
- Random selection:
- No
- Main weakness:
- Bias within groups
Convenience
- Uses groups:
- No
- Random selection:
- No
- Main weakness:
- Often unrepresentative
Which method uses groups but does not choose members at random?
Errors and missing data
Never silently keep an impossible value: check it, correct or remove it, and report the problem.
Data may contain recording errors, such as typing mistakes, impossible values like 400 hours in a week, or a stopwatch left running. Data may also be missing.
Missing data matter most when the missing items differ from the rest, because then the remaining data are biased.
- 1
Check the source
and correct the value if you can
- 2
Remove it
if it cannot be corrected
- 3
Follow up
non-respondents, or replace them
- 4
Report the issue
and treat conclusions with caution
A student records 400 hours of revision in one week. What first?
Outliers
A value is an outlier if it is more than below or above .
An outlier is below or above , where . The distance is measured from the nearest quartile.
In context an outlier may be genuine (an exceptionally fast runner) or an error, so check before removing. Outliers appear as crosses on box and whisker diagrams (4.2) and inflate the range and standard deviation (4.3).
Worked example
Test the data for outliers.
- 1
Quartiles: , , so .
- 2
Upper boundary: . Lower boundary: .
- 3
, so is an outlier. No value is below .
, . What is the upper boundary for outliers?
Try an exam question
A school has 500 students: 200 in Year 10, 180 in Year 11 and 120 in Year 12. A stratified sample of 50 is taken. Find the number from each year, and state one advantage of stratified over convenience sampling.
[4 marks]
- [1]Year 10: .
- [1]Year 11: .
- [1]Year 12: .
- [1]Advantage: every year group is represented in proportion, so less bias.
That's the notes covered.
Carry on to the next subtopic.
Try a question
Which sampling technique has the researcher used?
1 mark, marked against the mark scheme
Which sampling technique has the researcher used?
1 mark, marked against the mark scheme