Source
Adapted from my own handwritten and hand-typed notes from Year 1 to Year 3 of my EEE degree
A confidence interval for the population mean has the form , where is the margin of error.
Requirement
must be approximately normal:
- if , assume the population is approximately normal
- if , the Central Limit Theorem gives this for free
Case 1: sigma known (Z interval)
Find from Table A1 so that . Rearranging the inequality for :
Common values: (90%), (95%), (99%).
Case 2: sigma unknown (t interval)
Replace by the sample standard deviation . Then
has a t distribution with d.f. It is symmetric with mean 0, but flatter with heavier tails than . As grows it approaches . is the value with area to its right (Table A3).
If the d.f. is too large for the table, use instead.
Method
Steps:
- Check that is approximately normal
- Find the critical value or (with d.f.)
- Compute the margin of error
- Write the interval and state it in words
Worked example (sigma known)
Survey scores with . A sample of gives . Find a 95% interval.
Solution:
- , so CLT applies, and
- interval:
Worked example (sigma unknown)
, , , population assumed normal. Find a 95% interval.
Solution:
- with 15 d.f. is 2.131
- interval:
Choosing the sample size
From , the smallest sample for a required margin is
and we round up. The same critical values are used to build rejection regions in a Hypothesis Test for the Mean.