BIAS II - Bayesian methods for integrated bias modelling and analysis of multiple data sources in observational studies

Principal Investigator: Professor Nicky Best

BIAS II (Bayesian methods for integrated bias modelling and analysis of multiple data sources in observational studies) focused on addressing methodological challenges in the modelling of biases and complex structure in observational data, in particular surveys, longitudinal studies and small area data. These methodological challenges include data linkage and methods which combine data from different databases; methods for modelling bias due to survey non-response; longitudinal and spatial analysis, including extension of multilevel models to handle multiple time series and improved methods for handling spatial autocorrelation, scale issues and estimating distributions of small area statistics.

Researchers at BIAS II used Bayesian graphical and hierarchical models as a natural tool for linking many different sub-models and data sources and accounting for important sources of heterogeneity and patterns of correlations in observational data. They applied these models to a range of social science problems covering topics such as crime, voting behaviour, ageing, health and policy evaluation using small area indicators.

BIAS II also ran a programme of training events to help build capacity in advanced quantitative methods – particularly Bayesian methods. Their training programme included specialist workshops, mentoring schemes and visiting fellowships, and they developed freely available software tools to implement many of our methods.

BIAS II was based at Imperial College London.

For further  information please see BIAS II website.