Date: Mon 21 March 2011
Duration: 1 day (10.30am - 4:00pm)
Level: Level 1 (Introductory)
Location: Oxford University - Lecture Theatre in the Manor Road Building ( http://www.manor-road.ox.ac.uk/index.php/finding-us-google-map.html)
Fee: This event is free of charge. (Please note that tea/coffee will be provided in breaks, but lunch is not provided. There is a canteen nearby where lunch can be purchased, or you may wish to bring your own lunch with you).
Registration: No pre-registration is required
Tutors: Nicky Best, Steve Fisher, Jane De Lance Holmes
This workshop provides an introduction to a new class of multilevel models – termed hierarchical related regressions (HRR) – for estimating individual-level associations using a combination of aggregate (group level) and individual-level data. HRR differs from other methods by enabling analysts to model individual and aggregate data simultaneously, while including information on the dependent variable at the aggregate level (e.g. constituency election results), and data from aggregation units not available at the individual level (e.g. census data from all constituencies or output areas in the country). The workshop will also discuss HRR as a method of improving ecological inference (analyses that aim to make inference on the relationship between individual-level quantities using aggregate data). The HRR models combine features of standard ecological regression models for aggregate data and multilevel models for clustered individual-level data, and have been shown to reduce bias and improve precision in many situations.
Lectures will cover the following
Quantitative social scientists who wish to analyse data individual-level data linked with data aggregated to a geographical (or other group) unit of analysis. This includes analyses where only aggregate data are available (ecological inference) and analyses of individual-level data where aggregate characteristics (such as proportion of ethnic minority people in the area/group) are included as explanatory variables.
Good working knowledge of ordinary least squares and logistic regression
Nicky Best is Professor of Statistics and Epidemiology at Imperial College School of Public Health, London. She is Director of the BIAS node of the ESRC National Centre for Research Methods, and has research interests in Bayesian statistical modelling of health and social science data. She regularly teaches short courses on applied Bayesian methods and spatial epidemiology.
Stephen Fisher is a Lecturer in Political Sociology at the University of Oxford. He has research interests in electoral behaviour, social attitudes and research methods. He teaches quantitative methods for social scientists and ran the Oxford Spring School for Quantitative Methods in Politics and Sociology from when it began in 2003 to 2010.
Jane De Lance Holmes is a Research Associate in the Department of Epidemiology and Biostatistics at Imperial College London. Her current research focuses on development and application of ecological and hierarchical related regression models for understanding electoral behaviour and for explaining small area variations in chronic diseases.