School of Mathematical Sciences Queen Mary, University of London Autumn 2001 STATISTICS SEMINAR: DESIGN OF EXPERIMENTS All are welcome The talks are held at 16.30, all in the Mathematics Seminar Room (103) on Level 1 of the Mathematics Building, Queen Mary, University of London. Tea and coffee are available in the Mathematics Common Room (102) from 15.00. The nearest underground station is Stepney Green. Turn left at the exit and walk 400 yards. ________________________________________________________________________ DATE SPEAKER TITLE ------------------------------------------------------------------------ 11 Oct 2001 M.A. Ollis Protection Against Premature Queen Mary, Termination of Experiments Based University of London on Williams Squares With Circular Structure 18 Oct 2001 T.E. O'Brien Robust Experimental Design Loyola University Chicago, Strategies for Dose Response Katholieke Universiteit Models Leuven 15 Nov 2001 S. Lewis and D. Woods Designing Experiments for University of Southampton Semi-Parametric B-Spline Models 6 Dec 2001 B. Jones and S. Bate Universally Optimal GlaxoSmithKline Cross-Over Designs when the Number of Subjects and Periods is a Multiple of the Number of Treatments. ----------------------------------------------------------------------- For more information ask: Barbara Bogacka School of Mathematical Sciences Queen Mary, University of London Mile End Road London E1 4NS Tel: 020 7882 5497 e-mail: [log in to unmask] --------------------------------------------- The seminar information is kept on: http://www.maths.qmw.ac.uk/~rab/seminars.html ______________________________________________________________________ A B S T R A C T S ______________________________________________________________________ M.A. Ollis Protection Against Premature Termination of Experiments Based on Williams Squares With Circular Structure It is known that Williams squares with circular structure are E-optimal in certain circumstances. We introduce a particular type of Williams square with circular structure called a witch-hat square which may be preferable to other Williams squares with circular structure if it is possible that the experiment may be terminated prematurely. We show that witch-hat squares exist for all n, where n is different from 4k, give some enumeration results for small $n$ and report that a computer search has shown that there are no witch-hat squares of order 8 (it is known that there are no Williams squares with circular structure of order 4). ------------------------------------------------------------------------- T.E. O'Brien Robust Experimental Design Strategies for Dose Response Models In the context of nonlinear regression models, this paper outlines recent developments in design strategies when the assumed model function, initial parameter guesses and/or error structure are not known with complete certainty. Designs obtained using these strategies are termed robust designs as they are intended to be resistant to specified departures. Robust designs are clearly advantageous in many practical settings since these designs can be used to test for, say, lack of fit of the assumed model function or for error heteroskedasticity, whereas so-called optimal designs often cannot. --------------------------------------------------------------------------- S. Lewis and D. Woods Designing Experiments for Semi-Parametric B-Spline Models The work discussed in this talk was motivated by the need to design factorial experiments in industry for a response which may have several turning points or exhibit non-smooth behaviour. Flexible and parsimonious semi-parametric models for such a response will be defined which are built from B-spline basis functions and monomials. Prior information on anticipated behaviour of the response can be incorporated into the model via choice of the location and the degree of smoothness of the knots. An advantage of the models is that model-fitting and inferential methods for general linear models can be applied. An algorithmic approach to finding efficient designs under these models will be outlined. The key issue of uncertainty in the locations of the knots at the design stage will be addressed through two extensions of standard design search criteria. These approaches will be compared and contrasted over a variety of examples. ---------------------------------------------------------------------------- B. Jones and S. Bate Universally Optimal Cross-Over Designs When the Number of Subjects and Periods is a Multiple of the Number of Treatments Abstract will be given later. ----------------------------------------------------------------------------