Home | Stats Matter | Toolkit | Statistical Sampling

Statistical Sampling

Stats Matter

Statistical Sampling

Aim of this topic

To provide basic guidance on issues to consider when designing a sample.

Using a survey to collect information

A major aspect of any research is the gathering of information. Surveys are an effective tool to collect information directly from a source and conduct research to use in decision making. Surveys collecting information from all of a population (complete enumeration) are referred to as a Census. Sample surveys collect information from only part of the population (partial enumeration).

There are a number of alternatives to using surveys. Information may already be available from another source for example from administrative by-product data 1 from a previous survey or from existing research. Case studies can enable the collection of very detailed information while participatory observation methods are often used to collect qualitative information and to examine group dynamics and explore people’s perceptions.

Non-survey methodologies can provide information, often very detailed data, at a relatively low cost and in a short time frame. However there may also be disadvantages, statistical accuracy may be lower and existing data sources may not match information requirements very well.

The following resources provide further information on the use of surveys to collect information. 


Census or sample?

A census provides a true measure of a population (no sampling error) and can allow detailed information to be collected about small sub groups within a population, however they can be expensive and take a long time to conduct. Sample surveys tend to have lower costs and are more timely however data collected is often not suitable for producing benchmark data and is subject to sampling error. However, if good sampling techniques are used, results can be representative of the actual population.

The following resources provide more information around the advantages and disadvantages around sample surveys and statistical sampling.



Sampling methods

There are a number of considerations when designing a sample: cost, level of accuracy required and timing. Depending on  requirements a sample can be random (probability sampling) or non-random (non-probability sampling).

Probability or random sampling enables population estimates to be produced and estimates of sampling error to be produced. Common approaches include simple random sampling, systematic sampling and stratified random sampling. 

Under non-probability or non-random sampling methods estimates of sampling error cannot be produced, making it difficult to infer estimates of the population from the sample. Common approaches include convenience sampling, purposive sampling, quota sampling and volunteer sampling.

The following resources provide information around sampling techniques and designs commonly used. The main focus is on determining an appropriate sampling method.


Sample size

Determining the size of your sample is a very important step in statistical sampling and can impact the quality of the data obtained. The sample size refers to the number of individuals or groups required to respond (not just the number approached) to achieve the required level of survey accuracy. There are no fixed rules for deciding on the sample size, factors to consider include the population size and variability, incidence of characteristics being measured in the population, sample design, resources, level of accuracy required, likely level of non-response and sampling methods used.

The following resources provide information around calculating a sample size.


Sampling and non-sampling error

Sampling errors occur when data are collected from a sample rather than a census. It refers to the difference between an estimate for a population based on data from a sample and the true value for that population that would result if a census were taken. Sampling error can be measured and controlled in random sampling techniques.

Non-sampling error can occur in both sample surveys and a census. It is caused by factors other than those related to sample selection and can occur at any stage of a survey or census. Examples of non-sampling error include processing errors, interviewer errors, response errors and non-response errors.

The following resources provide information sampling and non-sampling error.



1 Data collected by government agencies and service providers as part of their case management, clinical or other administrative records about clients and the nature of their transactions with the agency or service provider.

Document key
 HTML page
 Link to external site
 PDF file
 MS Word
 MS Powerpoint
 MS Excel