Events

Date: 5 March 2026   Time: 14:00 - 15:00

Location: Hybrid: MB503, SMS, QMUL, or via the Teams link below

I will describe a new class of flexible mixed-effects models for simple clustered or longitudinal data. The idea of these Adaptively-Structured Mixed Models (AdaStruMMs) is to replace pre-specified random-effects structure by unknowns to be estimated. AdaStruMMs meet three requirements, at least one of which is unmet by all existing approaches I have considered:

1. Sufficient flexibility to represent individual trajectories as arbitrary smooth curves.

2. Information sharing between individuals to exploit structural similarities.

3. Accurate estimation of individual trajectories and population summaries across different data structures.

I will demonstrate the high-quality inference offered by AdaStruMMs through a range of simulations, and give some ideas about visualising and interpreting model fit through an example on changes in body composition in adolescent girls.

Teams link