Precision and Bias
Increasing precision and reducing bias are the two primary objectives in
conducting a reliable survey:
- Precision is about how precise or accurate your
results are; how narrow the confidence intervals are
for estimates from your survey
- Bias is about whether your survey really is
representative of the population of residents.
For example, a survey might give an estimate of
smoking prevalence in the adult population in an area as
15%. With 1,000 respondents this could be accurate to
within +/- 3% (a precise estimate!). But, if the sample
excludes younger adults, the estimate of smoking
prevalence will be biased (even if it is accurate!).
This is an extreme case – but more commonly seen
where some sub-groups of the population have lower
response rates than others.
For an excellent easy-to-read summary of sampling
errors, precision, bias and confidence intervals, see
Hoinville et al pages 56-60.
Further information on
Non-response Bias, including research on it's effects is
available by clicking