Royal Statistical Society
Meeting of the RSS Social Statistics Section
24 April 2007, 5.00pm at the Royal Statistical Society (refreshments from
4.30pm)
ONS Investigation of GREG Based Calibration Weighting for Non-Response:
Case Study Using the Labour Force Survey.
Chair: James Brown ([log in to unmask])
Speakers: Paul Smith (ONS) and Peter Brodie (ONS)
Discussant: Prof Chris Skinner (University of Southampton)
Is it worth attending?
Any one who uses the weights with social survey data (or is involved in
their production) should find this meeting informative!
Outline
For many years now, ONS has used calibration weighting as part of its
adjustment for unit non-response on Social Surveys. This has been
implemented using 'raking ratio' and is also commonly applied to ad hoc
surveys commissioned by Government Departments that are carried out by a
variety of market research and survey organizations in the UK. From this
session you will see why we use calibration weighting for non-response and
an alternative method of implementing it.
The first part will give a general overview of why we use calibration
techniques to adjust for non-response. This will help users to see when and
how calibration weighting helps with non-response.
The second part will then look at the methods of implementation. In
particular, it will consider a GREG based approach as an alternative to
raking ratio methods. The advantage of this approach is that the model
behind the calibration is explicit rather than implicit. A case study of the
implementation using the Labour Force Survey demonstrates how the method
performs in practice.
The meeting will take place at the Royal Statistical Society headquarters,
12 Errol Street, London, EC1Y 8LX (see
http://www.rss.org.uk/main.asp?page=1759 for map and directions). There is
no charge for this event and pre-registration is not required.
|