Error types

Error is inherent in all analyses. It is crucial to be aware of these errors, and to design your analysis in such a way that it minimises any bias in your data, or likelihood that your data is compromised because of methodological error. The Australian Bureau of Statistics website(ABS){: .link-ext target=”_blank” } describes two categories of error types when undertaking an analysis:

  • sampling error - the sample does not represent the population
  • non-sampling errors - systematic or random errors in the sample

There are many types of non-sampling errors, such as:

  • coverage error - data is included or excluded in the sample
  • non-response error - the failure to obtain a response
  • response error - respondents intentionally or accidentally providing inaccurate responses
  • interviewer error - incorrect recording of data at the point of collection
  • processing error - error as a result of inputting or coding

Example applications

It is important to be careful in the selection of wording when asking people about an issue. The ABS notes that the wording of questions can lead to non-sampling errors, such as questions that rely on memory recall, ask for confirmation of a socially desirable behaviour, encourage under- or over-reporting, are leading, or are double-barrelled.

For example, and without generalising, having a parent ask an engineering student how much alcohol they can drink in a night could lead to under-reporting, whereas having a classmate ask the same question might lead to over-reporting. If you ask the same student how much time they volunteer per week, it might be artificially inflated, as volunteering is a socially desirable pursuit.

Tips and techniques

When designing your survey or approach, there are a few principles that can assist:

  • keep your survey short and targeted
  • use neutral language to avoid bias
  • assure the interviewee of confidentiality in the process
  • providing advance notice of the planned activity
  • ensuring the sample size is large enough to account for non-responses
  • triangulate your questions using multiple measures of similar things

Key concepts

  • an overview of the error types
  • an example of an incorrect or corrected survey, based on reducing the likelihood of error
  • advice to the student engineer on how to approach the design of a survey to reduce error

Core resources

Updated:  12 Mar 2018/ Responsible Officer:  Head of School/ Page Contact:  Page Contact