Hypothesis test for correlation
Studying this course? Keep your progress in one place.
Key facts
- The product moment correlation coefficient satisfies and measures linear association only.
- comes from a sample; the population correlation coefficient is .
- Hypotheses use : against , or .
- Reject if lies in the critical region given by the critical value.
- Conclude in context. Correlation does not imply causation.
Correlation coefficient
measures how closely points lie to a straight line, from to .
The product moment correlation coefficient satisfies .
- : all points on a line with positive gradient.
- : on a line with negative gradient.
- near 0: no linear association.
Find on your calculator. A perfect curve can still give near 0.
Which value cannot be a product moment correlation coefficient?
Sample and population
describes the sample; the test uses it to make a claim about the population value .
is calculated from a sample. The population correlation coefficient is (rho).
Hypotheses are always written in terms of , never .
- 1
Sample
calculate the product moment correlation coefficient r
- 2
Hypothesis test
compare r with the critical value for the sample size n
- 3
Population
decide whether the population correlation ρ differs from 0
Sample
- Data collected
- Gives
Population
- Whole group of interest
- Has
Which is a correctly written null hypothesis for no correlation?
Stating the hypotheses
; the alternative is one- or two-tailed depending on the question.
The null hypothesis is . The alternative hypothesis is (two-tailed), or (one-tailed).
Use one-tailed only if the direction is stated before looking at the data.
Two-tailed
- 'There is correlation'
One-tailed
- or
- Direction expected in advance
A teacher expects more revision to give lower anxiety scores. Which is correct?
Using critical values
Compare with the critical value for your sample size and level; reject if it is in the critical region.
Use the table for and the level. One-tailed: reject if critical value () or critical value (). Two-tailed: reject if critical value.
If is in the critical region, reject ; otherwise do not.
Worked example
, two-tailed, 5% level, critical value . The sample has .
- 1
, .
- 2
Compare with .
- 3
, so is in the critical region.
Two-tailed test, critical value , . What do you do?
Conclusion and causation
State the decision about and what it means in context, never claiming proof or causation.
Alternatively use a p-value: reject if it is less than the significance level, such as .
Write whether is rejected and what that means in context, e.g. 'significant evidence of positive correlation between fertiliser and yield'.
Correlation does not imply causation: another variable may affect both, or the cause may run the other way.
Can conclude
- Evidence of correlation at the stated level
Cannot conclude
- That one variable causes the other
- That the test 'proves' anything
Fertiliser use and yield show significant positive correlation. What can you conclude?
Try an exam question
A sample of 15 pairs gives . The critical value for a one-tailed test at 5% with is . Test whether there is negative correlation, and comment on causation.
[4 marks]
- [1], .
- [1]Compare with .
- [1], so reject ; there is evidence of negative correlation at the 5% level.
- [1]Correlation does not imply causation.
That's the notes covered.
Carry on to the next subtopic.
Try a question
Which pair of hypotheses should the student use?
1 mark, marked against the mark scheme
Which pair of hypotheses should the student use?
1 mark, marked against the mark scheme