Computational Statistics & Data Analysis 45 (2004) 423 – 429 www.elsevier.com/locate/csda Construction of new skew Hadamard matrices and their use in screening experiments S. Georgiou, C. Koukouvinos ∗ , S. Stylianou Department of Mathematics, National Technical University of Athens, Zografou 15773, Athens, Greece Received 15 January 2002; received in revised form 26 February 2003 Abstract Screening experiments aim to identify the relevant variables within some process potentially depending on a large number of variables. Recently a new class of experimental designs called edge designs has been introduced. These designs allow a model-independent estimate of the set of relevant variables, thus providing more robustness than traditional designs. Some new skew Hadamard matrices of order 36 are constructed and used to produce new edge designs. An illustrative simulated example using an edge design in 70 runs is also presented. c  2003 Elsevier B.V. All rights reserved. Keywords: Skew Hadamard matrix; Edge design; Linear model; Construction 1. Introduction Screening experiments are used for the correct identication of relevant variables out of a large number of variables that possibly inuence some characteristic y. The aim of screening is to detect cheaply this subset of inuential variables; then more accurate designs on lower-dimensional subspace are used to account for nonlinearities. Since the noninuential variables are not explored in subsequent measurements, an important requirement for screening designs is that they place measurements such that the region of interest is well explored. The number of trials N needed for the identication of the active factors should be kept as small as possible. More specically, screening is addressed to estimate a multivariate function y such that y i ≃ y(x i ); x i =(x i 1 ;:::;x i n ) ∈ [a 1 ;b 1 ] ×···× [a n ;b n ]; * Corresponding author. Tel.: +30-210-7721706; fax: +30-210-7721775. E-mail address: ckoukouv@math.ntua.gr (C. Koukouvinos). 0167-9473/$ - see front matter c  2003 Elsevier B.V. All rights reserved. doi:10.1016/S0167-9473(03)00058-6