Edexcel 9MA0 · Paper 3, Statistics

Edexcel A Level Maths large data set: the weather data

Pearson Edexcel's large data set is a spreadsheet of daily weather readings from the Met Office. Some Paper 3 statistics questions are set in its context, and knowing the data set gives you an advantage: which places and months it covers, what each variable means, and how missing or tiny values are recorded.

Try the practice questions Get the data set from Pearson ↗

5 practice questions, 22 marks
Full worked solutions
Link to the official data set

What it is

The Edexcel large data set in brief

Examined in Paper 3 (Statistics and Mechanics), Section A: statistics. Always use the copy from Edexcel's own website: we don't host the data set, and boards can issue new versions.

Daily weather data provided by the Met Office.
Eight weather stations: five in the UK (Camborne, Heathrow, Hurn, Leeming and Leuchars) and three overseas (Beijing, Jacksonville and Perth).
Two periods: May to October 1987 and May to October 2015. That is 184 days per station per year.
Pearson says the data set should be appropriate for the lifetime of the qualification and is reviewed each year, so check you have the current issue.

Variables and units

What is in the Edexcel data set

Column headings as they appear in the spreadsheet. The overseas stations have fewer variables recorded than the UK stations: check which columns are filled for each place in your copy.

Variable	Units	What to know
Daily mean temperature	°C	The mean temperature for the day. A continuous variable.
Daily total rainfall	mm	A trace (less than 0.05 mm) is recorded as tr.
Daily total sunshine	hours	Hours of sunshine in the day.
Daily mean windspeed	knots, and a Beaufort-scale description	The Beaufort conversion turns the speed into words such as 'light' or 'moderate'.
Daily maximum gust	knots	The highest gust recorded that day.
Daily maximum relative humidity	%	Relative humidity is a percentage of the air's capacity for water vapour.
Daily mean total cloud cover	oktas	Eighths of the sky covered: 0 is clear, 8 is completely overcast.
Daily mean visibility	decametres (Dm)	1 Dm = 10 m.
Daily mean pressure	hPa	Hectopascals.
Wind and gust direction	degrees and compass (cardinal) points	Direction the wind blows from.

Codes and conventions

tr	Trace of rain: less than 0.05 mm. For calculations, treat it as 0 and say that you have.
n/a	Not available: the reading is missing. Leave it out of calculations and reduce n (the sample size) to match.

For the exam

What you are expected to know

Where the stations are: Camborne (Cornwall, near the coast), Heathrow (London), Hurn (near Bournemouth), Leeming (North Yorkshire) and Leuchars (Fife, Scotland, on the coast). Overseas: Beijing (China), Jacksonville (Florida, USA) and Perth (Western Australia).
Perth is in the southern hemisphere, so its May to October data are autumn, winter and early spring there. The other stations are in the northern hemisphere, where May to October is late spring, summer and autumn.
The data only cover May to October. You can't use them to say anything about winter in the UK.
How to handle 'tr' and 'n/a' when cleaning data, and what the units mean (oktas, knots, Dm, hPa).
How to take simple random, systematic, stratified, quota and opportunity samples from the spreadsheet, and the difficulties of each with this data.
Which variables are continuous (temperature, rainfall) and which are recorded on a scale (oktas for cloud cover, the Beaufort description for windspeed).

Typical question styles

Describe how to take a sample of days from one station, for example a systematic sample.
Clean an extract: deal with tr and n/a, then find a mean, median or standard deviation.
Interpret a scatter diagram, correlation coefficient or regression line for two weather variables, and test for correlation.
Judge a probability model against the data, for example a discrete uniform model for cloud cover (as in 9MA0/03, June 2018).
Use your knowledge of the data set to comment on a claim: the months covered, the hemisphere, the location of a station.

Common mistakes

Treating tr as missing, or n/a as 0. tr is a real, tiny amount (use 0); n/a is no reading (leave it out and reduce n).
Forgetting that Perth's May to October is its winter half of the year.
Extrapolating a regression line beyond the range of the data, for example predicting humidity for more sunshine than any day in the sample had.
Writing a conclusion without context: say what the result means for the weather at that station.
Mixing up units: windspeed is in knots, visibility in decametres, not kilometres.

Practice

Edexcel large data set practice questions

Original questions in the style of the Edexcel papers, each with a worked solution and mark scheme. The numbers in these questions are made up for practice, in the style of the data set. They are not values from the real data set, so don't quote them in an exam.

Question 1Cleaning rainfall data5 marks

The daily total rainfall, in mm, at a UK weather station was recorded for 10 consecutive days:

0.0 tr 2.4 11.2 tr 0.3 n/a 5.8 0.0 1.6

State what tr means. [1]
Explain how you will deal with tr and n/a, then calculate the mean daily rainfall. [3]
The median of these data is 0.3 mm. Give a reason why the median may be a better average than the mean here. [1]

Worked solution and mark scheme

tr means a trace of rain: less than 0.05 mm fell. B1
Treat each tr as 0 mm. n/a means no reading, so leave that day out: 9 values remain. B1
Total $= 0 + 0 + 2.4 + 11.2 + 0 + 0.3 + 5.8 + 0 + 1.6 = 21.3$ mm. M1
Mean $= 21.3 \div 9 = 2.37$ mm (3 s.f.). A1
The data are skewed: one very wet day (11.2 mm) pulls the mean up, while most days had little or no rain. The median is not affected by that extreme value. B1

Question 2A systematic sample of days4 marks

Amira wants a systematic sample of 15 days from the data for Leeming for May to October 2015.

Explain how she should take her sample. [2]
On one of her chosen days the variable she is studying is recorded as n/a. State what she should do. [1]
Explain why she cannot use her sample to draw conclusions about the weather at Leeming in January. [1]

Worked solution and mark scheme

May to October has $31 + 30 + 31 + 31 + 30 + 31 = 184$ days. Number the days 1 to 184. $184 \div 15 \approx 12.3$, so use an interval of 12. M1
Choose a random start between 1 and 12, then take every 12th day after it, 15 days in all. A1
Leave that day out of the calculation (and say the sample is now 14 days), or replace it with the next available day. B1
The data set only covers May to October, so it has no January data; conclusions can't be extended to other months. B1

Question 3Modelling cloud cover4 marks

In an extract of the data set, daily mean total cloud cover is recorded as a whole number of oktas from 0 to 8. Ben models the cloud cover $X$ on a randomly chosen day with a discrete uniform distribution over the values 0, 1, 2, …, 8.

Using Ben's model, find $P(X \ge 6)$. [2]
For 30 days at one station, the cloud cover was:

Oktas	0	1	2	3	4	5	6	7	8
Days	1	1	2	2	3	4	6	8	3

Use these data to comment on Ben's model. [2]

Worked solution and mark scheme

Each of the 9 values has probability $\tfrac{1}{9}$. M1
$P(X \ge 6) = P(6) + P(7) + P(8) = \tfrac{3}{9} = \tfrac{1}{3}$. A1
From the data, the proportion of days with 6 or more oktas is $\tfrac{6 + 8 + 3}{30} = \tfrac{17}{30} \approx 0.567$. M1
This is much larger than the model's $\tfrac{1}{3}$, and the frequencies are far from equal (1 day with 0 oktas, 8 days with 7), so a discrete uniform model is not suitable for these days. A1

Question 4Sunshine and humidity: correlation7 marks

For a random sample of 12 days at a UK station, Chloe recorded the daily total sunshine $s$ hours and the daily maximum relative humidity $h$ %. The values of $s$ ranged from 0.4 to 11.8. She found the regression line of $h$ on $s$:

$h = 94.2 - 2.35s$

and the product moment correlation coefficient $r = -0.81$.

Interpret the gradient of the regression line in context. [1]
Estimate the maximum relative humidity on a day with 6 hours of sunshine. [1]
Explain why the line should not be used to estimate $h$ on a day with 15 hours of sunshine. [1]
Test, at the 5% significance level, whether there is evidence of negative correlation between sunshine and maximum relative humidity. The critical value for a one-tailed test with $n = 12$ at 5% is $-0.4973$. [4]

Worked solution and mark scheme

For each extra hour of sunshine, the maximum relative humidity falls by about 2.35 percentage points. B1
$h = 94.2 - 2.35 \times 6 = 80.1$ %. B1
15 hours is outside the range of the data (0.4 to 11.8 hours), so this would be extrapolation and may not be reliable. B1
$H_0: \rho = 0$, $H_1: \rho < 0$, where $\rho$ is the correlation coefficient for all days at this station. B1
Compare: $-0.81 < -0.4973$, so $r$ is in the critical region. M1
Reject $H_0$. A1
There is evidence, at the 5% level, of negative correlation between daily sunshine and maximum relative humidity at this station. A1

Question 5Knowing the data set2 marks

Explain why the May to October data for Perth come from a different season from the May to October data for Heathrow. [1]
Dan says: "The data set shows that UK winters are getting milder." Explain why the data set can't support this claim. [1]

Worked solution and mark scheme

Perth is in the southern hemisphere, so May to October is autumn, winter and early spring there, but late spring, summer and autumn at Heathrow. B1
The data set only covers May to October (in 1987 and 2015), so it contains no winter data for the UK. B1

Keep going with Edexcel statistics

Sampling and Data Presentation Probability Statistical Distributions Hypothesis Testing Statistics practice Edexcel past papers Edexcel A Level Maths

Other boards' data sets: AQA 7357 (cars data) · OCR A H240 (census data) · All large data sets

Facts about the data set were checked against Edexcel's published data set and guidance in September 2026. If anything here differs from the version your teacher gives you, the board's version is right. A Level Math Revision is independent and is not affiliated with Pearson Edexcel.

Find your weakest statistics topic in 5 minutes.

Free, no sign-up. Then try every topic for 7 days, no card needed.

Take the free diagnostic