Modelling impact heterogeneity

    Research output: Contribution to journalArticlepeer-review


    The treatments embodied in social interventions are characterized by their heterogeneity, delivered as they often are by different individuals operating in different social and geographical contexts. One implication of this heterogeneity is that average treatment effects will often be less useful than estimates of differential impacts across contexts. The paper shows how multilevel models can be used to estimate variability of impact and to account for systematic effects. These models are specified for multisite interventions, for studies using cluster allocation and for designs that incorporate matching. The paper indicates how qualitative and quantitative approaches to evaluation could be linked.
    Original languageEnglish
    Pages (from-to)31-38
    Number of pages7
    JournalJournal of the Royal Statistical Society. Series A: Statistics in Society
    Issue number1
    Publication statusPublished - 2002


    • Area-based initiatives
    • Cluster trials
    • Formative evaluation
    • Multilevel models
    • Randomized controlled trials
    • Summative evaluation


    Dive into the research topics of 'Modelling impact heterogeneity'. Together they form a unique fingerprint.

    Cite this