Hypothesis tests for a binomial proportion
Studying this course? Keep your progress in one place.
Key facts
- Define the parameter , then write using the value claimed, never the sample proportion.
- Under the number of successes is .
- The p-value is for and for .
- In a two-tailed test compare the tail probability with half the significance level.
- Conclude in context: reject , or say there is insufficient evidence. Never 'prove'.
Setting up the test
Define , write the hypotheses in terms of , and state the distribution of under .
Define the parameter (for example the proportion of all parcels delivered on time), then write against .
The sample gives an inference about the population. If is true, the successes are .
The significance level is the probability of incorrectly rejecting when it is true.
- (the claimed value)
- If is true
A company claims 85% of parcels arrive on time. In a sample of 15, 10 do. Which is the correct ?
One-tailed test
Find the p-value in the direction of and compare it with the significance level.
- Define and write and .
- State the distribution of under .
- The p-value is for or for .
- Compare with the level and conclude in context.
- 1
Hypotheses
define ; and
- 2
Distribution
under
- 3
p-value
or
- 4
Conclusion
compare with the level, in context
Worked example
A drug cures 30% of patients. 10 of 20 patients on a new drug are cured. Test at 5% whether the new drug is better.
- 1
Let be the probability a patient is cured. , .
- 2
Under , .
- 3
-value .
- 4
, so reject .
For and observed , which probability is the p-value?
Two-tailed tests
When , split the significance level equally between the two tails.
Compare the tail probability with half the significance level, after deciding the tail from the mean .
For the mean is 12, so 17 is in the upper tail: . At 5% compare with : , so do not reject .
Alternatively double the tail probability and compare with the full level.
| One-tailed | Two-tailed | |
|---|---|---|
| Tail probability compared with | The full level | Half the level |
| Split of the level | None | Equal in each tail |
One-tailed
- Tail probability compared with:
- The full level
- Split of the level:
- None
Two-tailed
- Tail probability compared with:
- Half the level
- Split of the level:
- Equal in each tail
Two-tailed test at 5%: a one-tail probability is 0.02. Do you reject ?
Critical region method
Find the critical region first, then see whether the observation lies in it.
Choose the critical region as large as possible with probability at most the significance level. If the observed value is in it, reject .
The actual significance level is the probability of that region under .
Worked example
, , 1% level. Find the critical region and decide whether 6 is significant.
- 1
, so is too large.
- 2
, so the critical region is .
- 3
The actual significance level is .
- 4
6 is not in the critical region.
What is the actual significance level in the worked example?
Writing the conclusion
State what happens to , then say what it means in context, without claiming proof.
A full conclusion has two parts.
- Reject : 'There is evidence at the 5% level that the new drug cures more than 30% of patients.'
- Do not reject : 'There is insufficient evidence at the 5% level that the proportion on time is less than 85%.'
Failing to reject does not prove the claim; the sample may be too small.
Reject
- Evidence for at this level
- Say what it means in context
Do not reject
- Insufficient evidence for
- Never 'accept' or 'prove'
Which is the best conclusion when is not rejected?
Try an exam question
A seed company claims 80% of its seeds germinate. A gardener plants 25 seeds and 16 germinate. Test at the 5% level whether the proportion is lower than claimed, given for .
[5 marks]
- [1], , where is the proportion that germinate.
- [1]Under , .
- [1]p-value .
- [1], so reject .
- [1]There is evidence at the 5% level that fewer than 80% of seeds germinate.
That's the notes covered.
Carry on to the next subtopic.
Try a question
State the distribution of if the new drug is no more effective than the old one.
1 mark, marked against the mark scheme
State the distribution of if the new drug is no more effective than the old one.
1 mark, marked against the mark scheme