Confidence Intervals
Surveys are rarely 100% accurate. At best, they
produce ‘precise’ estimates of population valuesFor
example, a lifestyle survey might find that 25% of
adult residents in an area currently smoke. This
figure is an estimate of the true (unknown)
proportion of smokers in the area – the actual
(exact) proportion who smoke could only be found by
a census of all residents.
Given that surveys produce estimates, we
give confidence intervals for the estimates to show
how accurate they are. For example, for our survey
estimate of the number of smokers (25%), the 95%
confidence interval for this estimate might be 22.8%
to 27.2%. Here, we are 95% confident that the true
proportion of smokers is between 22.8% and 27.2%.
The value 95% is the confidence level.
Other confidence levels commonly used are 90% and
99%.
For SPSS data analysis for this example
click here
|
