The value of survey data is entirely dependent on the quality of sampling. Whenever a sample is drawn from a population, the data relating to that sample will contain a degree of error. This error is known as sampling error.
Generally, the larger the sample size, the more accurate the data and the more accurate the conclusions about the whole population. For large populations (over 1,000) it is the absolute size of a sample which determines sampling error, not the relative size of the sample compared with the population from which it is drawn.
The target sample size for a survey depends on three main factors:
- the resources available
- the aim of the study
- the statistical quality needed for the survey.
For qualitative surveys using focus groups or interviews, the sample size needed will be smaller than if quantitative data is collected by questionnaire.
If statistical analysis is to be performed on the data then sample size calculations should be conducted.
Below is an example of a useful site to use for sample sizing:
that will allow event organisers to enter some key pieces of information and calculate the sample size needed to achieve a certain level of confidence in the results, e.g. the site calculator Raosoft suggests that for example for an event attended by 20,000 visitors, with a margin of error of 5% and confidence level of 95%, recommended survey sample is 377 visitors.
Sampling error only applies to truly random samples, i.e. those which are representative of the populations from which they have been drawn.
In light events or festivals located at city centre, it is almost impossible to demonstrate that truly random sampling has been carried out. At many events there is only a limited window of opportunity to collect data from those attending.
Consequently, the pragmatic approach is to survey as large a sample as possible within the time and resources available (convenience sampling) in the hope that is represents the population as reasonably as possible and that the size of the sample minimises any potential biases.