Seminars and Events
All welcome (no need to pre-register)
All seminars take place in Keppel Street, London WC1E 7HT. External visitors will be met at reception 10 minutes prior to the start time.
For further information please contact Jonathan Bartlett (jonathan.bartlett@lshtm.ac.uk) or Ruth Keogh (ruth.keogh@lshtm.ac.uk)
Where permission is given by the speaker, seminars will be audio recorded with slides, and made available on this page.
Seminar Programme, Summer 2015
Thursday, 21 May, 12.45pm, LG80
Maarten J. Bijlsma (University of Groningen, The Netherlands)
"Bias assessment of a causal inference approach to age-period-cohort analysis"
Wednesday, 27 May 2015, 12:45pm, LG80
Prof. Alan Agresti (Professor Emeritus, University of Florida)
"Some Perspectives about Generalized Linear Modeling"
Slides with audio
Wednesday, 22 July 2015, 12:45pm, LG9
Dr. Aidan O'Keeffe (University College London)
"Dynamic Causal Inference in a Regression Discontinuity Design using Local Independence"
Regression discontinuity (RD) designs have been developed as method for causal inference from observational data, through the exploitation of a naturally occurring intervention rule. Most RD design methods have focussed on causal inference where the outcome of interest is continuous. Scenarios in which the outcome of interest is binary have received less attention.
A regression discontinuity design consists of three processes: an intervention process, a binary outcome process and a threshold attainment process (where the decision on whether or not to apply an intervention is taken based on the attainment of a fixed ‘threshold’ by a subject-specific continuous ‘assignment variable’). These three processes may change at different points over time and we examine the dynamic relationships amongst the processes and consider how associated causal relationships may be inferred.
Properties of local independence are used to develop a suitable framework in which the dependencies amongst these processes may be modelled over time, under particular assumptions, and causal inference made concerning the effect of a previous intervention on the time until the occurrence of a binary outcome.
More seminars to be added to the programme later
Previous seminars with recordings
Wednesday, 25 February 2015, 12.45pm,
Dr Andrew Simpkin (MRC Integrative Epidemiology Unit, School of Social and Community Medicine, Bristol)
"Estimating features of repeated measures trajectories"
