The Applied Statistics and Data Science (ASDS) research group brings together statisticians in the School of Engineering, Physics and Mathematics who develop and apply statistical methods to problems in health, medicine and social science.
A
central strand of our research is life-course and global health epidemiology,
underpinned by causal inference methods. We are interested in how circumstances
early in life, such as childhood socioeconomic position, education and the
economic conditions people grow up in, shape health and cognitive ageing
decades later. We also study how these patterns differ between high-income
countries and low- and middle-income settings, where we have long-standing
collaborations, particularly in Indonesia. Questions like these can rarely be
answered by experiment, so much of our methodological effort goes into drawing
credible causal conclusions from observational data. That means instrumental
variable and Mendelian randomisation approaches, careful handling of missing
data and selection, and building and validating risk prediction models,
increasingly with machine learning alongside classical methods.
Alongside
this, the group has real methodological breadth. Much of the data we work with
changes over time, varies across places, or both. We have particular expertise
in longitudinal and time-to-event data, including joint models that link
repeated measurements to survival outcomes. We use Bayesian hierarchical models
for spatial and spatio-temporal problems such as disease mapping. This work
extends to infectious disease and environmental surveillance, including models
that combine wastewater monitoring with randomised prevalence surveys to track
COVID-19 at national scale. We also develop methods for data that break
standard distributional assumptions, centred on finite mixture models fitted by
the EM algorithm. These cover model-based clustering and classification of
skewed, heavy-tailed, incomplete and matrix-variate data, robust regression and
factor analysis, and models for censored outcomes and measurement error. A
further strand develops geometric, non-Euclidean statistical methods for data
that live on trees and networks, from phylogenetics to graph-valued data in
genetics and neurology. Our analyses are carried out mainly in R and Stata, and
we put weight on reproducible work: our code is written so that others can
check, reuse and build on our methods.
Current
projects include dementia risk prediction in low- and middle-income settings;
socioeconomic inequalities in health across the life course using international
ageing cohorts (ELSA, SHARE, HRS and IFLS); early markers of Parkinson’s
disease from muscle activation patterns, with colleagues at Newcastle
University; and statistical leadership on NIHR-funded clinical trials and
service evaluations. Our work has been funded by NIHR, MRC, NHS trusts and
industry partners, and has appeared in The Lancet, The Lancet Global Health,
American Journal of Epidemiology, JRSS Series A, Computational Statistics and
Data Analysis and Systematic Biology.
Our
team
- Dr Eduwin Pakpahan (Group Leader), Associate Professor. Medical Statistics and Epidemiology, Causal
inference in observational data, life-course health inequalities, dementia risk
prediction, and longitudinal methods.
- Dr Guangquan Li (Deputy Group Leader), Assistant Professor.
Bayesian methods for health and social science data, spatial and
spatio-temporal modelling, and disease surveillance, including wastewater-based
monitoring of COVID-19.
- Dr Maryam Garba, Assistant Professor. Geometric and
non-Euclidean statistics for tree- and graph-valued data, including
phylogenetic trees, with applications in neurology such as early detection of
Parkinson’s disease.
- Dr Mehrdad Naderi, Assistant Professor. Finite mixture models
and the EM algorithm for clustering, classification and factor analysis, with a
focus on skewed and heavy-tailed distributions, censored and incomplete data,
and statistical computation.
We
welcome enquiries from prospective PhD students, and from researchers,
clinicians and organisations looking for statistical collaboration. We are also
happy to advise on the statistical side of research grant development, from
study design and sample size to analysis plans, ideally at an early stage
rather than close to the deadline.
For
statistical advice or to discuss a possible collaboration, please contact the
Group Leader, Dr Eduwin Pakpahan, at <Eduwin.Pakpahan@northumbria.ac.uk>
