Hypothesis test for the parameter of a geometric distribution
Studying this course? Keep your progress in one place.
Key facts
- Observe one value of , the number of trials to the first success, and test .
- and : no tables needed.
- A larger means a shorter wait: uses ; uses .
- Solve critical regions with logarithms, reversing the inequality when dividing by .
- The test assumes independent trials with constant , and one value gives a fairly weak test.
The geometric test
You observe one value of and decide whether it is surprising if has its stated value.
A geometric distribution counts the independent trials up to and including the first success.
In a hypothesis test for , you observe one value of , the test statistic, and decide whether it is surprising if has its stated value.
Tail probabilities need a single power, with no tables. Remember a larger makes smaller: a small observed is evidence that is higher, and a large evidence that is lower.
For , what is ?
Stating the hypotheses
Write and an alternative about the parameter , never about .
The null hypothesis is , where is the stated probability. The alternative hypothesis is , or , depending on the suspicion described.
Define in context, for example "the probability that a card wins a prize". Then state that under , .
Writing is a common error: the hypotheses concern the parameter , while is the observed value.
p lower
- One-tailed
- Expect a longer wait
p higher
- One-tailed
- Expect a shorter wait
p different
- Two-tailed
A game claims a chance of winning. A player thinks it is lower. Which alternative hypothesis is correct?
One-tailed tests
Choose the tail by asking which direction is "more extreme", then compare with the significance level.
Assume is true. Find the probability of a result at least as extreme as the observed and compare it with the significance level.
For a lower means a longer wait, so use . For a higher means a shorter wait, so use .
- Longer wait is extreme
- Shorter wait is extreme
Worked example
A card game claims of winning a prize. The first prize comes on card . Test at the level whether is less than .
- 1
, . Under , .
- 2
.
- 3
Compare: , so do not reject .
Worked example
A screen is defective with probability . The first defective screen is the nd tested. Test at the level whether is greater than .
- 1
, . Under , .
- 2
.
- 3
Compare: , so reject .
and the first success is on trial . Which probability do you compare with the significance level?
Critical regions
Solve the tail inequality with logarithms, round to a whole number and check the neighbouring values.
The critical region is the set of values of that lead to rejecting .
For it has the form , where is the smallest value with the significance level. Solve with logarithms, reversing the inequality because . Then check the neighbouring values.
For it has the form with . For this gives , since .
The actual significance level is the probability of the critical region under .
- , with
- , with
Worked example
under and at the level. Find the critical region and the actual significance level.
- 1
Solve . Taking logs: .
- 2
Dividing by reverses the inequality: , so .
- 3
Check: but , so .
- 4
The actual significance level is .
For at , . Is in the critical region?
Two-tailed tests
Put half the significance level in each tail and check each tail separately.
For , put half the significance level in each tail. At use in each tail.
- Lower tail : .
- Upper tail : .
Check each tail separately, and say clearly if one of them is empty.
Lower tail
Upper tail
Worked example
under and at the level. Find the critical region.
- 1
Lower tail: , so there is no lower critical region. The smallest possible value of is not extreme enough.
- 2
Upper tail: but , so and .
- 3
The actual significance level is .
In the example above, why is there no lower critical region?
Conclusions and assumptions
State whether is rejected, then conclude in context, and comment on independence and constant .
State whether is rejected, then give a conclusion in context: "There is sufficient evidence, at the level, that the probability of a defective screen is greater than ." If is not rejected, say there is insufficient evidence against the claim. Never say a claim is proved or disproved.
The test relies on independent trials with a constant probability of success. If the probability changes, for example because a player improves with practice, the geometric model is not suitable.
A test based on a single observed value is also fairly weak.
- 1
Hypotheses
State and in context
- 2
Distribution
Write under
- 3
Direction
Decide which tail is extreme
- 4
Probability
Use a single power, no tables
- 5
Conclude
Reject or not, then a sentence in context
Which assumption is needed for the geometric model?
Try an exam question
A game claims that each play wins with probability , independently. A player thinks the probability is lower. The player's first win comes on the th play. Test the claim at the significance level.
[5 marks]
- [1], .
- [1]Under , .
- [1].
- [1].
- [1]Reject : there is sufficient evidence, at the level, that the probability of winning is less than .
That's the notes covered.
Carry on to the next subtopic.
Try a question
Which of the following gives the correct hypotheses, where is the probability that a card wins a prize?
1 mark, marked against the mark scheme
Which of the following gives the correct hypotheses, where is the probability that a card wins a prize?
1 mark, marked against the mark scheme