Chhetri AcademyGCSE & A level Paper Builder

S5.1, S5.3Tests for correlation and for the mean of a normal distribution

Edexcel A level Maths (9MA0) · Statistics › Statistical hypothesis testing

Practise Tests for correlation and for the mean of a normal distribution. 3 exam-style questions plus unlimited generated ones on this subtopic, at up to four difficulty levels, with full mark schemes and a progress tracker. Free, no account needed.

Build a paper on this topic

▶ Watch videos on Tests for correlation and for the mean of a normal distribution (TLMaths on YouTube) · Practise all of Statistical hypothesis testing

Sample questions

Written for this site in the style of Edexcel exam questions. They are not taken from real past papers.

Question 1Easy2 marks
A researcher collects data on the time spent on a phone each evening and the number of hours slept that night for a random sample of teenagers. The researcher wants to test whether there is negative correlation between the time spent on a phone each evening and the number of hours slept that night.
Write down suitable hypotheses for the test, defining the parameter you use.[2]
Show the answer and mark scheme
Answer: \(H_0: \rho = 0, \ H_1: \rho \lt 0\), where \(\rho\) is the population product moment correlation coefficient between the time spent on a phone each evening and the number of hours slept that night.
  • B1 for \(H_0: \rho = 0\) with \(\rho\) the population correlation coefficient
  • B1 for \(H_1: \rho \lt 0\)
Question 2Medium7 marks
A teacher records the number of hours of revision and the test mark for a random sample of 18 students. The product moment correlation coefficient between hours of revision and test mark is \(r = 0.42\).
Critical values of the product moment correlation coefficient for a sample of size 18 are 0.4000 (one-tailed test at 5%) and 0.4683 (one-tailed test at 2.5%, or two-tailed test at 5%).
(a) Test, at the 5% level of significance, whether there is evidence of positive correlation between hours of revision and test mark. State your hypotheses clearly.[4]
(b) A second teacher tests the same data at the 5% level for evidence of correlation, without specifying positive or negative. State the hypotheses and the conclusion of this test.[2]
(c) Explain why the two conclusions do not contradict each other.[1]
Show the answer and mark scheme
(a) Answer: \(H_0: \rho = 0\), \(H_1: \rho \gt 0\); \(0.42 \gt 0.4000\), so reject \(H_0\): there is evidence of positive correlation between hours of revision and test mark.
  • B1 for \(H_0: \rho = 0\), \(H_1: \rho \gt 0\)
  • M1 for comparing 0.42 with 0.4000
  • A1 for reject \(H_0\) (significant)
  • A1 for a conclusion in context

Worked solution: \(H_0: \rho = 0\), \(H_1: \rho \gt 0\), where \(\rho\) is the population correlation coefficient. Critical value 0.4000. Since \(0.42 \gt 0.4000\), reject \(H_0\): there is evidence of positive correlation between hours of revision and test mark.

(b) Answer: \(H_0: \rho = 0\), \(H_1: \rho \ne 0\); \(0.42 \lt 0.4683\), so do not reject \(H_0\): insufficient evidence of correlation.
  • B1 for \(H_1: \rho \ne 0\)
  • B1 for \(0.42 \lt 0.4683\), not significant: insufficient evidence of correlation between hours of revision and test mark

Worked solution: \(H_0: \rho = 0\), \(H_1: \rho \ne 0\). Critical value 0.4683 (two-tailed, 5%). Since \(0.42 \lt 0.4683\), do not reject \(H_0\): there is insufficient evidence of correlation.

(c) Answer: They test different alternative hypotheses: a one-tailed test puts all 5% in one tail, so it needs weaker evidence in that direction. The result is borderline, significant for one test but not the other.
  • B1 for the tests have different alternative hypotheses (one-tailed puts all of the 5% in the upper tail, so its critical value is lower); the evidence here is borderline

Worked solution: The first test only looks for positive correlation and puts the whole 5% in the upper tail, giving a lower critical value. The second allows for correlation in either direction, so it needs stronger evidence in each direction. With a borderline value such as 0.42, one test can be significant and the other not.

Question 3Hard9 marks
A machine fills bottles with water. The volume, in ml, of water in a bottle is normally distributed with mean \(\mu\) and standard deviation 4 ml. The machine is set so that \(\mu = 500\).
After the machine is serviced, a random sample of 25 bottles has a mean volume of 498.3 ml.
(a) Test, at the 5% level of significance, whether there is evidence that the mean volume has decreased. State your hypotheses clearly.[5]
(b) A student writes: “The test proves that the machine now under-fills, and its new mean volume is 498.3 ml.”
Give two criticisms of this statement.[2]
(c) Explain whether the conclusion would be different if the test had been two-tailed at the 1% level of significance.[2]
Show the answer and mark scheme
(a) Answer: \(H_0: \mu = 500\), \(H_1: \mu \lt 500\); \(\bar{X} \sim \mathrm{N}(500, 0.8^2)\), \(P(\bar{X} \le 498.3) = 0.0168 \lt 0.05\): reject \(H_0\). There is evidence that the mean volume of water in the bottles has decreased.
  • B1 for \(H_0: \mu = 500\), \(H_1: \mu \lt 500\)
  • M1 for \(\bar{X} \sim \mathrm{N}\left(500, \frac{4^2}{25}\right)\)
  • M1 for \(P(\bar{X} \le 498.3)\) or \(z = \frac{498.3 - 500}{0.8} = -2.125\)
  • A1 for 0.0168 (or \(z = -2.125 \lt -1.6449\), or critical value 498.68)
  • A1 for a conclusion in context: reject \(H_0\); there is evidence that the mean volume has decreased

Worked solution: \(H_0: \mu = 500\), \(H_1: \mu \lt 500\). Under \(H_0\), \(\bar{X} \sim \mathrm{N}\left(500, \frac{16}{25}\right) = \mathrm{N}(500, 0.8^2)\).
\(P(\bar{X} \le 498.3) = P(Z \le -2.125) = 0.0168 \lt 0.05\).
Reject \(H_0\): there is evidence at the 5% level that the mean volume of water in the bottles has decreased.

(b) Answer: A test gives evidence, not proof (there is a chance of rejecting a true \(H_0\)); 498.3 ml is only the sample mean, an estimate of the new population mean.
  • B1 for a significant result is evidence, not proof: there is up to a 5% chance of rejecting \(H_0\) when it is true
  • B1 for 498.3 is the mean of this sample only; the population mean is unknown (498.3 is an estimate)

Worked solution: A hypothesis test cannot prove anything: with a 5% significance level there is a chance of getting such a sample even if \(\mu = 500\). Also 498.3 ml is the mean of this sample, which only estimates the new population mean; another sample would give a different value.

(c) Answer: Yes: two-tailed at 1% needs a probability below 0.005 in the lower tail (or \(z \lt -2.5758\)); \(0.0168 \gt 0.005\), so \(H_0\) would not be rejected.
  • M1 for comparing 0.0168 with 0.005 (or \(2 \times 0.0168 = 0.0336\) with 0.01, or \(-2.125\) with \(-2.5758\))
  • A1 for not significant: insufficient evidence that the mean volume has changed, so the conclusion would be different

Worked solution: For a two-tailed test at 1%, each tail has 0.5%. \(P(\bar{X} \le 498.3) = 0.0168 \gt 0.005\) (equivalently \(z = -2.125 \gt -2.576\)), so \(H_0\) would not be rejected: there would be insufficient evidence that the mean volume has changed.

Related subtopics

Stuck? Get 1-to-1 help. Chhetri Academy tutors GCSE and A level Maths and Science online, with a free 30-minute trial lesson.

Book a free trial