Supersaturated Latin Hypercube Designs

Abstract

Latin hypercube designs have received much recent interest in the context of computer experiments where there may be many input variables (or factors). Supersaturated designs offer an eficient method of determining the factors that have the most substantial effect on the response of interest. In this article, supersaturated Latin hypercube designs are constructed and they are shown to have a number of favorable properties including near-E(s2)-optimality. They are also shown to facilitate a straighrforward analysis in terms of the spectral properties of the designs.

Block-circulant matrices for constructing optimal Latin hypercube designs

“…The proposed constructions can also be expanded and used for building supersaturated Latin hypercube designs as these were initially introduced in Butler (2005). Moreover, it would be interesting to investigate if the known methods for constructing orthogonal Latin hypercube designs with 2 c + 1 runs and 2 c factors can be modified and used for the construction of low-correlation designs.…”

Nearly orthogonal latin hypercube designs for many design columns

“…Efthimiou, Georgiou, and Liu (2014) and some other references constructed nearly orthogonal LHDs by adding runs to the existing LHDs, while we accommodate more design columns while keeping the same run size and preserving near orthogonality. This is important as computer experiments often look at more factors than existing orthogonal LHDs can afford (Butler (2005)).…”

Design of Experiments for Screening

“…This latter application is the most important for screening where the resulting data could be used to estimate a surrogate linear model with dummy variables or a GP model with an appropriate correlation structure [90]. Supersaturated LHS designs, for d ≥ n, have also been developed [21].…”