Design of Experiments

Definition of design of experiments

Aspects of scientific experiments such as samples size, computing power, time and financial overhead, are often limited by the availability of resources.  The design of experiments relates to making decisions under uncertainty about the allocation of these resources, such as the number of experimental units required, the allocation of subjects to treatments and the allocation of available resources. The aim of experimental design is to aid statistical inference about the underlying process that generates the observed data, within the bounds of the available resources.

Who works in design of experiments?

Students: Amy Cook, Liz Ryan
Post-docs:
Lecturers: James McGree, Chris Drovandi, Gentry White, Helen Thompson
Professors: Tony Pettitt

What expertise do we have?

  • Design of experiments under model uncertainty.
  • Design of experiments to aid parameter estimation.
  • Developing design of experiments methodology for high dimensional design space.
  • Developing design of experiments methodology for large numbers of observations within the design space.
  • Using Gaussian process smoothing as an efficient method to optimise experimental designs on a continuous space.
  • Using parallel processing using GPU, supercomputers to developing scalable algorithms for experimental design.

Case studies:

The design of experiments is fundamental to a wide range of research activities including the following:

Environment:

  • In collaboration with the Australian Institute of Marine Science, model-based adaptive monitoring methods are being developed for spatio-temporal coral reef data. Modelling allows updating of existing information using past information, so new decisions incorporate lessons learnt from past observations.

Medical:

  • Clinical trials: James McGree

Pharmacokinetics:

  • Drug development study where the design is the optimal drug dose and patient sampling times.Engineering: Design of experiments under model uncertainty for chemical reaction models.

Engineering:

  • Design of experiments under model uncertainty for chemical reaction models.

Biology:

  • Design of experiments for stochastic epidemiological models such as modelling the evolution of macro-parasites. Macro-parasites, typically transmitted by mosquitoes, cause lymphatic filariasis disease in an estimated 120 million people worldwide. One approach is to use indirect inference to avoid calculation of an intractable likelihood.

Contact

Amy Cook: a20.cook@connect.qut.edu.au

Example:

5 Big Data Capability_ Design of Experiments example

Source: Lindenmayer et al. Adaptive monitoring in the real world: proof of concept. Trends in Ecology & Evolution, 26(12):641−646, 2011

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s