Surveys - Not Just for People

The word survey means different things to different people. To a land developer it means measuring out blocks of land. To a builder it is a review of a structure’s quality. To a social researcher it usually means a collection of information from people. To an academic it may be a review of the knowledge in an area.

The common element of these definitions is the obtaining of an overview – a definition that matches the origin of the word, the Old French sourveoir, which in turn comes from the Latin words super (above) and videre (to see).

To a statistician a survey means gathering information from a population.  The use of the word population betrays the origin of statistical surveys in understanding people.  Early in the development of statistics it was realised that it was often not necessary to collect information on everyone (a census) but just looking at a subset of the population could give reasonable accuracy if chosen properly.  This is a sample survey, but the aim is still to obtain an overview of the whole population.  The statistician’s expertise covers the methods of choosing the sample, a blend of theory and practical experience.

The principles of surveys and sampling developed for people can be applied in many other contexts, even though they are not always called surveys.  For example, in quality management it is not efficient to inspect every single product coming off a production line, but it may be possible to inspect a small subset and on that basis decide whether the production line is operating satisfactorily.  Accountants use sampling of transactions when conducting an audit – checking a sample may provide sufficient confidence that the vast majority of transactions are accurate and hence the financial statements based upon them are a true and fair view. 

These areas have developed Standards for sampling.  In quality assurance a set of Standards developed by the American military (MILSPECs) are commonly applied for this purpose.  Australian Accounting Standards include a standard for sampling.  However, the adaptation of survey methods to a new application requires an understanding of the application area, and a careful consideration of the statistical principles that underlie the methods.  The Standards rarely explain these statistical principles, so a statistician is required to determine how best to apply them, even if the departure from a previous application may seem to be minor.

Data Analysis Australia regularly designs sampling processes for surveys in applications as diverse as inventory validation (where a survey may replace a complete stocktaking), environmental impact studies and quality assessment as well as, of course, people.  These sampling processes frequently give clients substantial savings since the alternatives can be very labour intensive.  Good design through statistics means that these are truly cost effective solutions – lower cost and highly effective.

March 2014