Overview
Topic status: We're looking for students to study this topic.
Project Summary
Optimal experimental design methodology is important in many applied fields such as medicine, biology, finance and computer science. Optimal designs lead to experimental aims being achieved in a timelier manner, hence saving on monetary resources and potentially reducing the burden on those involved in the experiment. The majority of optimal experimental design approaches rely on the statistical model being of a sufficiently simplistic form so that the probability model (or likelihood function) can be evaluated point-wise. However, there is an ever-increasing need to specify more complicated models in order to mimic reality more closely. The aim of this research is to make use of and extend the available literature on likelihood-free inference, which has received considerable recent attention for parameter estimation, and experimental design in order to design experiments for models with computationally prohibitive likelihoods.
The project could involve application of methodology to interesting applications in biology, for simplification.
Expected outcomes, applications and/or benefits
The research will give rise to novel experimental design algorithms as well as the development of innovate design utility functions where likelihood evaluations are not required. This research will extend the applicability of optimal experimental design to a wider class of statistical models and thus broaden the fields to which this methodology can be applied.
The research can be applied to various models with biological or epidemiological applications. The project could focus solely on this applied work.
Required student skills/experience
- At least Statistical Modelling 2 (MAB314).
- Statistical Inference (MAB524) also preferred.
- Programming skills: Matlab and/or C or an equivalent language.
- Study level
- Vacation research experience scholarship
- Supervisors
- QUT
- Organisational unit
Science and Engineering Faculty
- Research area
- Keywords
- experimental, design
- Contact
- Contact the supervisor for more information
Dr Christopher Drovandi
