National Statistics Methodology Advisory Committee
First meeting, 31 October 2001

The following work in progress papers were considered:

Paper 1: Overview of methodological work within the Government Statistical Service (GSS)
The Methodology Group (MG) provide methodological support and advice on the statistical processes used in the Office for National Statistics (ONS). This paper presents Methodology Group's vision and direction, plans and key objectives.

Committee Conclusions:
Members agreed that research should have a higher priority in MG business objectives. More thought should be given to collaboration with external experts, and to the dissemination of methods and results of research. Increasing the quantity and quality of publications will increase awareness of MG outside ONS, which is important for recruitment purposes. Government departments should be supplied with information on the structure of MG, so that they know whom to contact when faced with a specific methodological problem.

Paper 2: Small Area Estimation within the ONS
This paper describes the work carried out to estimate Local Authority level unemployment measured according to the International Labour Organisation definition.

Committee Conclusions:
The Committee raised a number of issues requiring further consideration, including the use of:

• narrower age-sex bands and additional covariate data;
• calibration to alternative margins;
• random effects;
• time series methods;
• spatial smoothing;
• breakdowns in the diagnostic tests.

Paper 3: Estimating variance of movements in time series data: issues arising for the Average Earnings Index (AEI)

This paper discusses the estimation of standard errors of movements for the UK Average Earnings Index, but most of the issues raised would be relevant for any set of time series data.

Committee Conclusions:
Suggestions to follow up include:

• the reference to Yates, F. (1949) Sampling Methods for Censuses and Surveys;
• the use of standard modelling methods (eg Maximum Likelihood Estimation, the Expectation and Maximisation algorithm);
• using positive definiteness as a criterion for validity (an nxn matrix A is 'positive definite' if v'Av>0 for all non-zero n-vectors v);
• removing trend from the data before estimating covariances;
• estimating covariances for changes in levels;
• giving thought to ideas given on presentation of standard errors.

Paper 4: Recruitment, development and retention of methodologists within the ONS

This paper outlines the major issues facing ONS and the strategy being developed for the recruitment, development, and retention of methodologists.

Committee Conclusions:
Internships, joint appointments, joint projects with schools and universities, seminars involving external experts and sabbaticals should be considered as recruitment strategies for methodologists. MG work can be highly specialised and research-based, but many tasks are more general and more aptly described as requiring 'expert' administrative skills. Consideration must be given to the ways of efficiently partitioning these aspects into work packages, which will motivate and inspire staff with varying interests.

Second meeting, 13 March 2002

This page last revised: Tuesday, 28 May 2002