Hypothesis testing for zero correlation
In a nutshell
Based on the product moment correlation for a given sample, r, you can predict that measure for the population, ρ, using a hypothesis test.
One-tailed test
This is used to test if ρ is greater than zero or less than zero (i.e. testing for positive correlation or testing for negative correlation). One of two cases can be tested:
- H0:ρ=0;H1:ρ>0
- H0:ρ=0;H1:ρ<0
Two-tailed test
This is used when you only want to test if ρ is not equal to 0 (i.e. testing whether there is any correlation at all):
- H0:ρ=0;H1:ρ=0
Critical region
The critical region for r for your hypothesis can be determined using a given table of critical values. This region will depend on the significance level and the sample size:
- The bigger the significance level, the smaller the critical value.
- The bigger the sample size, the smaller the critical value.
- The smaller the critical value, the larger the critical region. This is because for a critical value c, the corresponding critical region is ∣r∣>c.
Example 1
For a given sample of 8 and a significance level of 10%, if H1:ρ>0, the critical region is r>0.5067. This means that:
- If the sample r<0.5067 , there is not enough evidence to reject H0.
- If the sample r>0.5067 , you can reject H0.
Example 2
The marks on two different subjects were taken for 30 students, which produced a product moment correlation coefficient of −0.51. Test, with a significance level of 10%, whether or not there is a negative correlation between the marks in these two subjects for all the students.
For the given sample and a significance value of 10%, you have:
H0:ρ=0H1:ρ<0
You also have, by the table, r<−0.3061.
Because −0.51<−0.3061, you can reject H0.
There is evidence, at the 10% level of significance, that a greater mark in the first subject means a smaller one in the second.