Testing for zero correlation

Testing for zero correlation

Spec 5.2 · Edexcel International A Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • A sample coefficient can be non-zero by chance, so test whether the population coefficient ρ\rho is non-zero.
  • H0:ρ=0H_0:\rho=0. The tail of H1H_1 comes from the wording of the question.
  • Compare rr (or rsr_s) with the critical value from the table, not with the significance level.
  • Not rejecting H0H_0 never proves there is no correlation.

Why test a correlation

A sample can show a non-zero value by chance, even when the population has no correlation.

A sample correlation coefficient comes from only some of the population. A hypothesis test decides whether the sample value is far enough from 0 to be convincing evidence of correlation in the population.

The population correlation coefficient is ρ\rho (product moment) or ρs\rho_s (Spearman). Hypotheses are about the population parameter, never about the sample correlation coefficient rr or rsr_s.

  1. 1

    Population

    Correlation coefficient ρ, usually unknown

  2. 2

    Sample taken

    Gives r or rₛ from the data

  3. 3

    Hypothesis test

    Is r far enough from 0 to be evidence about ρ?

From population to sample and back.

Sample

  • rr or rsr_s
  • Calculated from the data
  • Varies from sample to sample

Population

  • ρ\rho or ρs\rho_s
  • Fixed, usually unknown
  • What the hypotheses are about

Which is a correct null hypothesis?

Hypotheses and tails

The direction comes from the wording of the question, not from the sign of the sample value.

H0:ρ=0H_0:\rho=0 (no correlation). For a two-tailed test, H1:ρ≠0H_1:\rho\ne0, when the question asks whether there is any correlation. For a one-tailed test, H1:ρ>0H_1:\rho>0 or H1:ρ<0H_1:\rho<0, when the question states a direction in advance.

Use the same pattern for Spearman's, with ρs\rho_s.

−1−0.8−0.6−0.4−0.200.20.40.60.81-1H₀: ρ = 01

Two-tailed

  • H1:ρ≠0H_1:\rho\ne0
  • "any correlation"

One-tailed positive

  • H1:ρ>0H_1:\rho>0
  • "positive correlation"

One-tailed negative

  • H1:ρ<0H_1:\rho<0
  • "negative correlation"

A researcher asks whether there is a positive correlation between hours revised and mark. What is H1H_1?

The product moment test

Use random pairs from a bivariate Normal distribution; compare rr with the critical value for nn.

Use this test when the data are random pairs from a bivariate Normal distribution, so the relationship is linear. Find rr, find the critical value for nn and the significance level from the tables, and compare.

Two-tailed: reject H0H_0 if ∣r∣≥|r|\ge the critical value. One-tailed positive: reject if r≥r\ge it. One-tailed negative: reject if r≤−r\le- it.

−1−0.8−0.6−0.4−0.200.20.40.60.81-0.63190r = 0.620.6319

Worked example

n=10n=10, r=0.62r=0.62. Test for correlation at the 5%5\% level (two-tailed). The critical value is 0.63190.6319.

One-tailed negative test, critical value 0.50.5. Sample r=−0.6r=-0.6. What is the decision?

Spearman's test

Rank the data, find rsr_s, and compare with the critical value for nn.

Use this test when the data are ranks, when the relationship is monotonic but not necessarily linear, or when the data are not bivariate Normal. Rank the data, find rs=1−6∑d2n(n2−1)r_s=1-\frac{6\sum d^2}{n(n^2-1)}, and compare with the tabulated critical value, using the same rules.

For n=8n=8 the 5%5\% critical values are 0.64290.6429 (one-tailed) and 0.73810.7381 (two-tailed). For n=10n=10 they are 0.56360.5636 and 0.64850.6485.

00.20.40.6n = 8n = 10Sample sizeCritical value
  • One-tailed
  • Two-tailed
5% critical values of Spearman's rank coefficient from the note: two-tailed values are larger, and values fall as n increases.

Product moment test

Use when:
Linear relationship
Statistic:
r
Assumption:
Bivariate Normal

Spearman's test

Use when:
Ranks, or monotonic relationship
Statistic:
r_s
Assumption:
None about the distribution

Worked example

n=8n=8 and ∑d2=26\sum d^2=26. Test at the 5%5\% level (a) for positive correlation, (b) for any correlation.

Which test should you use for data that are ranks from two judges?

Writing the conclusion

Say "sufficient" or "insufficient evidence of correlation", in context.

If the sample value is in the critical region, reject H0H_0: "there is sufficient evidence of (positive/negative) correlation between ... and ...". If not, do not reject H0H_0: "there is insufficient evidence of correlation". Never say this proves there is no correlation: the sample may be too small to detect one.

A significant result shows evidence of correlation, not that one variable causes the other.

  1. 1

    Compare r with the critical value

    Is it in the critical region?

  2. 2

    If yes, reject H₀

    Sufficient evidence of correlation

  3. 3

    If no, do not reject H₀

    Insufficient evidence of correlation

  4. 4

    Write it in context

    Correlation does not show cause

From the test statistic to the conclusion.

In the critical region

  • Reject H0H_0
  • Sufficient evidence of correlation

Not in the critical region

  • Do not reject H0H_0
  • Insufficient evidence of correlation

A test does not reject H0H_0. Which conclusion is correct?

Try an exam question

A researcher claims that there is negative correlation between the price of a product and the number sold. For 1010 shops the product moment correlation coefficient is r=−0.71r=-0.71. Test the claim at the 5%5\% level. The one-tailed critical value for n=10n=10 is 0.54940.5494.

[4 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

Which pair of hypotheses should the teacher use, where ρ\rho is the population product moment correlation coefficient?

1 mark, marked against the mark scheme

Revision notes