Designing and analysing a cluster randomised experiment: Principles and practicalities
Date:
05/05/2021
Organised by:
London School of Economics and Political Science
Presenter:
Dr Krisztian Posch
Level:
Intermediate (some prior knowledge)
Contact:
Camilya Maleh
methodology.research@lse.ac.uk
Venue: Online
Description:
'Police in the classroom' was a cluster- block-randomised experiment with one pre-treatment and two post-treatment surveys. Its primary goal was to assess whether the presence of a police officer could increase trust in the police and learning among pupils compared to teacher or control conditions. This workshop will start with a discussion of the design principles of block- and cluster-randomised studies and how to integrate the two. It will be argued, that block-randomised experiments should be considered the 'gold standard' and that they can mitigate some of the issues prevalent with cluster-randomised experiment. Adding a longitudinal element to these experiments could provide further inference regarding the treatment effects.
The workshop will demonstrate the different methodological approaches that could be used to analyse a similar study. First, we will consider the results as if they were from a traditional RCT. Second, we will analyse the block-randomised portion of the data. Finally, the longitudinal subsample will be scrutinised. It will be also discussed how to reconcile the differing findings and how the interpretation of each of these approaches provide varying insight into the treatment effect. The workshop will use R as the statistical software to demonstrate these methods, so basic knowledge of this software will be expected.
Topics covered include:
- Creating blocks for block-randomisation
- Using blocks for cluster-randomisation
- Integrating a longitudinal element into experiments
- Analysis of the results using RCT, block-randomisation, and longitudinal approaches
Cost:
£30
Website and registration:
Region:
Greater London
Keywords:
Discourse Analysis, Creating blocks for block-randomisation , Using blocks for cluster-randomisation , Integrating a longitudinal element into experiments , Analysis of the results using RCT, block-randomisation, and longitudinal approaches
Related publications and presentations: