|
|
|
Operationalising Research
Inferential Statistics
- Only random sampling entitles the use of inferential statistics
- Below is the result of an imaginary survey in which a random sample of an adult population were asked whether or not they smoked.
|
|
|
|
|
|
Men
|
Women
|
Totals
|
Smokers |
743
|
924
|
1667
|
Non-smokers |
625
|
820
|
1445
|
Totals |
1368
|
1744
|
3112
|
|
|
-
- Null hypothesis: there is no difference in the population between the proportions of men and women who smoke
- The table suggests that 54% of the men and 53% of the women are smokers
- This might be taken to indicate that men are (slightly) more likely than women to be smokers
- Question
- How likely is it that a difference of this magnitude will arise purely as a result of random error?
- If the probability of obtaining a difference of this magnitude is less than 0.05 (1 in 20) then, conventionally, we can claim that the results are statistically significant
- Questions
- Does it matter whether the sample of 3112 people are drawn from the population of the UK (approx. 55 million) or from The People's Republic of China (approx. 1300 million)?
- If we multiply all of the numbers by 10 (eg imagine we had taken a random sample of size 31, 120), is the answer to the previous question the same?
|
|
|
|
Men
|
Women
|
Totals
|
Smokers |
7430
|
9240
|
16670
|
Non-smokers |
6250
|
8200
|
14450
|
Totals |
13680
|
17440
|
31120
|
|
|
|
|
from: random sample
|
|
|
|
|
|
|
|
|