Confidence intervals for a Normal mean
Studying this course? Keep your progress in one place.
Key facts
- A confidence interval is a range likely to contain ; about of intervals built the same way contain it.
- With known the limits are ; for .
- Higher confidence or larger widens the interval; larger narrows it.
- Round a required sample size up.
- A interval matches a two-tailed test at the level.
What an interval is
It is a range of plausible values for the parameter, built from a sample.
A confidence interval is a range calculated from a sample that is likely to contain the unknown parameter. The confidence level (for example ) says how reliable the method is: if you built an interval from each of many random samples, about would contain the true parameter.
is fixed; it is the interval that changes from sample to sample. A higher level gives a wider interval, and a larger sample a narrower one.
- 95% confidenceabout 95% of intervals contain μ
A 95% confidence interval for a mean is (49.2, 54.8). Which statement is correct?
Limits with known variance
Use , where cuts off the central area.
If with known, then . The confidence limits are , where is the standard error and cuts off the required central area:
- :
- :
- :
- :
Worked example
, , . Find a confidence interval for .
- 1
Standard error: .
- 2
For , .
- 3
Limits: .
Which -value gives a 99% confidence interval?
Width and sample size
The width is ; solve for and round up.
The interval is symmetrical about , with half-width . The width falls if increases, and rises if the confidence level or rises.
For a required width : , and always round up. Working backwards: a interval with has and half-width , so .
- Width
- Sample size
Worked example
. Find the smallest so that a interval has width at most .
- 1
Set .
- 2
So .
- 3
Then .
Doubling the sample size makes the width of the interval:
Interpreting an interval
It is a set of plausible values; the confidence level describes the method, not one interval.
A confidence interval does not say that of the population values lie in it, and it does not give the probability that a fixed is in a particular interval. Different samples give different intervals, and even a correct method fails to capture about of the time at the level.
In context, say "we are confident that the mean journey time lies between and minutes". The interval is a set of plausible values.
Valid
- Repeated intervals: about 95% contain μ
- The interval is a set of plausible values
Not valid
- 95% of the population lies in it
- Probability 0.95 that μ is in this interval
Interval A is (50.5, 54.3) at 95%. A 99% interval from the same data would be:
Link with hypothesis tests
A interval matches a two-tailed test at the level: reject if it lies outside.
To test against : if lies inside the interval, do not reject; if outside, reject.
A interval contains , so at the level is not rejected; it does not contain , so is rejected. A one-tailed test at needs a different interval, so use this link only for two-tailed tests.
A 95% interval for is (49.2, 54.8). At the 5% level, a two-tailed test of would:
Try an exam question
The mass of a bag of flour is Normal with standard deviation g. A random sample of bags has mean mass g above the nominal value. (a) Find a confidence interval for the mean excess. (b) Comment on a claim that the mean excess is g.
[4 marks]
- [1]Standard error .
- [1] limits: .
- [1]Interval .
- [1] lies outside the interval, so there is evidence the mean excess is not g.
That's the notes covered.
Carry on to the next subtopic.
Try a question
What is the standard error of the sample mean?
1 mark, marked against the mark scheme
What is the standard error of the sample mean?
1 mark, marked against the mark scheme