In the spotlight - Complex sample design guide and workshop

Article dated: 14 January 2010

Standard commands in statistical software typically treat data as simple random samples. The vast majority of ESDS Government supported surveys, however, employ complex sample design features such as clustering or stratification.

Software commands designed for simple random samples do not take into account the statistical implications of complex sample design. An important consequence of this is that the magnitude of standard errors may be underestimated (or in some cases, overestimated). Techniques are available in statistical packages such as Stata or SPSS that incorporate complex design features into the analysis to take into account such effects.

ESDS Government has produced an introductory guide which provides an overview of the survey design features of ESDS Government datasets. It also provides information on analysing complex samples using Stata. Focusing on design-based approaches, practical examples are given using the Health Survey for England (HSE).

An introductory workshop on complex sample design was held on 20 July 2009 at the University of Manchester. The morning sessions focused on theory and statistical concepts involved in the design and the analysis of complex surveys. In the afternoon, hands on workshops introduced design-based approaches to survey estimation using the 'svy' suite of commands in Stata. Topics covered include producing basic descriptive statistics, sub-population analysis, estimating design effects, and basic multivariate analysis.

The workshop will be repeated on 27 May 2010 at the University of Manchester. Further information, including a programme and booking form, will be available soon.