Source
Adapted from my own handwritten and hand-typed notes from Year 1 to Year 3 of my EEE degree
To test we compare the sample mean with , scaled by its standard error. Which statistic we use depends on whether is known.
Choosing the statistic
| Situation | Test statistic | Distribution under |
|---|---|---|
| known | , Table A1 | |
| unknown | with d.f., Table A3 |
In both cases must be approximately normal: assume normal when , or use CLT when .
Critical values
Tail rules:
- upper tail : reject if statistic (or )
- lower tail : reject if statistic (or )
- two tail : reject if (or )
Worked example (Z test, lower tail)
A new procedure should cut the 5 minute checking time. 36 cars give , and . Test at .
Solution:
- , CLT applies
- ,
- gives
- , so the statistic is not in the critical region. No significant evidence that the time has been reduced
Worked example (t test, lower tail)
A maker claims the mean bill for a new TV is less than 20. A sample of 10 gives and . Test at .
Solution:
- , so assume is normal
- ,
- , with 9 degrees of freedom
- , so the critical region is
- , so reject : the mean bill is significantly less than 20
The spread of a process is tested in a similar way in Hypothesis Test for the Variance.