You are here

  1. Home
  2. Research
  3. Research groups
  4. Statistics
  5. Design of Experiments

Design of Experiments

Optimal Design of Experiments

In an experiment, data are collected to answer a research question. Often, resources are limited, and designing the experiment optimally/efficiently for achieving this aim can save experimental effort and thus reduce the cost while enabling researchers to draw reliable conclusions from the data. Optimal design of experiments is an area of Statistics where we rephrase the research question in statistical terms to find the (statistically motivated) optimality criterion that best reflects the aim of the experiment.

The area of optimal design of experiments complements the data analysis/inference side of Statistics. Before we can find an optimal/efficient design, we first need to establish the type of statistical analysis to be used on the experimental data to be collected. The methodology is not restricted to a specific application area, but can be used widely for experiments in, e.g., Chemistry, Engineering or Medicine.

Prof Stefanie Biedermann leads the research on design of experiments in the school. Some examples of her research topics are:

  • Optimal design of experiments with potentially missing responses (EPSRC grant EP/V00641X/1)
  • Optimal design of survival experiments with potentially censored responses
  • Modelling and designing mixture experiments (e.g. Formulation Chemistry)
  • Optimisation algorithms for design search
Chart

Illustration: The figure below shows an optimal design for a chemical experiment. We can see the optimal sets of time points to measure the concentration of a chemical compound during a reaction, and the optimal temperatures. Here the aim was to get the most accurate estimate of the reaction rate (activation energy and pre-exponential factor in the Arrhenius equation).

Aims and Objectives:

  • to stimulate and facilitate research into the theory and applications of (optimal) design of experiments
  • to foster collaborative research, where appropriate;
  • to draw attention to the large range of expertise the Statistics Group can call on in this area