LEMMA 3: Longitudinal Effects, Multilevel Modelling and Applications
Principal Investigator: Professor Fiona Steele
The LEMMA 3 node combined methodological and substantive research with software development and capacity building. LEMMA 3 was based at the Centre for Multilevel Modelling at the University of Bristol.
LEMMA 3 carried out methodological research in four strands:
Progress on LEMMA 3's methodological research in 2012-13 included
- a Bayesian estimation procedure for a multivariate mixed response model with missing data on responses or covariates
- multiple imputation for missing data has been used in probabilistic record linkage
- tutorials on population-averaged vs. subject-specific models for multilevel discrete responses and on methods and software for modelling within-subject variability over time
LEMMA 3 also had a collaborative project Using survey data to enhance administrative data for policy relevant research with PEPA node of NCRM
LEMMA 3 developed MLwiN, REALCOM and e-STAT software and wrote templates for model specifications, estimation, data preparation and post-estimation predictions to aid interpretation of the fitted model.
Progress on software development included
- templates for missing data and data linkage have been written for the new Stat-JR software system to allow other researchers to implement these methods, and this offers substantial computational gains over the REALCOM-Impute software. Stat-JR also provides interoperability with other statistical packages such as Stata, R, SPSS, MLwiN, WinBUGS and aML
- further development of runmlwin Stata command.
The three substantive research projects by LEMMA 3 were
Progress on substantive research in 2012-13 included
- work on the interrelationship between employment transitions and mental health using simultaneous equations modelling of British Household Panel Survey (BHPS); further work is underway to study physical functioning trajectories by gender and socio-economic status.
- project on the determinants and consequences of occupation stress using real-time call record data on NHS24 nurses (with the Aberdeen Health Psychology Group).
LEMMA 3 also had a collaborative project on E-books for causal modelling and missing data methods with PATHWAYS node of NCRM.
LEMMA 3 publications are available in NCRM EPrints.
For further information please see LEMMA 3 homepage.