A selection of previous research projects
- LEMMA 1 Learning environment for multilevel methodology and applications (I)
- Realcom: Developing multilevel models for realistically complex social science data
- Family: Methodologies for studying families and family effects
- Multilevel multiprocess models for partnership and childbearing event histories
- Applications and understandings of multilevel modelling in the social sciences
- Applications of advanced multilevel modelling methods for the analysis of examination data
- Bootstrapping Measurement Error in Multilevel Models
- Developing Graphical and Inferential Tools for Social Data Analysis
- Advanced Training Workshops in Multilevel Modelling
- Optimal multilevel models of school effectiveness: Comparative analyses across regions
September 1999 - February 2003
- Project director: Harvey Goldstein
- Project officers: William Browne, Ian Plewis, Jon Rasbash, Min Yang
This project extended existing work on multilevel modelling in three main areas:
- The project continued to develop methodological tools for the analysis of complex social data; as increasingly sophisticated methodology was developed in this area it was apparent that further work is necessary to properly model the complexity of data from social systems.
- The project built on the existing MLwiN software with its innovative graphical interface in order to make it accessible to users in ways which mirror their own understandings of their data and help them to gain further insights into data structures.
- The project developed a range of dissemination and training materials, largely using the internet to raise awareness and understanding in the social science community and outside.
Applications of advanced multilevel modelling methods for the analysis of examination data
January 1998 - March 2001
- Project director: Harvey Goldstein
- Project officers: Min Yang, Geoff Woodhouse
The research had two aims.
- To develop existing multilevel methodology to handle efficiently measurement errors and ordered categorical responses.
- To provide substantive conclusions from an analysis of a very large cohort of students with GCSE and A level examination results.
This work has been supported by Economic and Social Research Council through grant number R000237394. Data provided by Department for Education and Employment.
Progress from GCSE to A and AS level: institutional and gender differences, and trends over time
This paper has appeared in the British Educational Research Journal, Vol. 27, No. 3. 2001:
- Progress from GCSE to A and AS level paper (zip, 0.1 mb)
A sub-sample of the data analysed in the paper is provided in the form of an MLwiN worksheet, along with information on the worksheet:
- Progress from GCSE to A and AS level worksheet (zip, 0.5 mb)
- Progress from GCSE to A and AS level worksheet information (PDF, 11kB)
Multilevel ordinal models for examination grades
- Fielding, A., Yang, M., and Goldstein, H. (2003). Multilevel ordinal models for examination grades. Statistical Modelling, 3 (2): 127-153.
- Multilevel ordinal models for examination grades paper (PDF, 765kB)
- Multilevel ordinal models for examination grades dataset (zip, 0.9 mb)
Multivariate multilevel analysis of examination results
- Multivariate multilevel modelling of A level examination results (PDF, 104kB) (J. Royal Statist. Soc, 2002, A, 165, 137-154).
This explores modelling strategies for multivariate (examination) responses with non-randomly missing responses within a multilevel data structure.
Research project evaluation
The project was graded as outstanding.
January 2000 - November 2000
It is well-known that measurement error in dependent or independent variables can lead to misleading inferences in regression-type applications (including multilevel modelling). There have been developments which enable the question to be dealt with in some situations, but these are not widely available and deal with only a proportion of the possible measurement error mechanisms. This project will develop a methodology based on bootstrap resampling which would enable simple treatment of a wide range of measurement error mechanisms. This project is taking place in conjunction with the development of MLwiN, and it is aimed to operationalise the results in future releases.
October 1998 - August 1999
In recent years statistical methods have been developed for analysing complex social data known as multilevel models, these procedures have found wide use in the social sciences such as education, economics, demography, politics etc. With the increasing power of personal computers and the quality of graphic displays it is now possible to implement more powerful methods that are very computer intensive and require sophisticated levels of understanding from users. These procedures promise to provide new insights into complex data structures. The project aims to make it easier for social scientists to gain access to these procedures via the use of informative graphical interfaces for the specification and analysis of these complex structures. The project will develop the existing MLwiN software interface for this purpose and provide exemplar materials.
April 1997 - April 1999
Following the work of the multilevel models project under phase 1 of ALCD, phase 2 of ALCD has allocated money for five advanced training workshops in multilevel modelling. There are a large number of users of this technique who now require help in extending their skills to handle a greater complexity of data, spatial data, time series etc. Five 2 day workshops are to be held at the Institute of Education for up to 12 invited participants at each. The topic areas and principal organisers are as follows: Education - Harvey Goldstein Environmental Economics - Ian Langford Health services data - Alastair Leyland Political Science - Anthony Heath Demography - Ian Diamond.
October 1995 - September 1998
Note: some of the documents on this page are in PDF format. In order to view a PDF you will need Adobe Acrobat Reader