As part of the Combechem project (EPSRC GR/R67729), methods and algorithms have been developed for the design of efficient and economic experiments when the response is appropriately described by a generalized linear model. An important application is to experiments with a binary response, such as the formation, or not, of a salt in crystallography. The statistical methods are described in

Woods, D.C., Lewis, S.M., Eccleston, J.A. and Russell, K.G. (2006) Designs for generalized linear models with several variables and model uncertainty, Technometrics, 48, 284-292.

Computer algorithms for finding designs under the criterion described in Woods et al ( 2006) are available on this page. Note that some code optimization has taken place since the publication of Woods et al (2006), which has increased the computational efficiency of the simulated annealing algorithm. Full details are available in

Woods, D.C. (2006) Designing experiments for binary data via simulated annealing, Technical Report 391, School of Mathematics, University of Southampton, UK.

Algorithms

- Search algorithms

• Simulated annealing - documentation
• Exchange algorithm - documentation

- Assessment algorithms - readme

Also available are designs and data discussed in Woods et al (2006).

Designs

- from Section 5 [xls]

- from Section 6

• designs from Example 1 [xls]
• designs from Example 2 [xls]

Data from the example in Section 7 [xls]

Queries and comments should be directed to Dave Woods

---

Last Updated: 10 May 2006