Abstract:Experimental designs intended to match arbitrary target distributions are typically constructed via a variable transformation of a uniform experimental design. The inverse distribution function is one such transformation. The discrepancy is a measure of how well the empirical distribution of any design matches its target distribution. This chapter addresses the question of whether a variable transformation of a low discrepancy uniform design yields a low discrepancy design for the desired target distribution. The answer depends on the two kernel functions used to define the respective discrepancies. If these kernels satisfy certain conditions, then the answer is yes. However, these conditions may be undesirable for practical reasons. In such a case, the transformation of a low discrepancy uniform design may yield a design with a large discrepancy. We illustrate how this may occur. We also suggest some remedies. One remedy is to ensure that the original uniform design has optimal one-dimensional projection, but this remedy works best if the design is dense, or in other words, the ratio of sample size divided by the dimension of the random variable is relatively large. Another remedy is to use the transformed design as the input to a coordinate-exchange algorithm that optimizes the desired discrepancy, and this works for both dense or sparse designs. The effectiveness of these two remedies is illustrated via simulation.

Centered $L_2$-Discrepancy of Random Sampling and Latin Hypercube Design, and Construction of Uniform Designs

A New Class of Latin Hypercube Designs with High-Dimensional Hidden Projective Uniformity

Optimal Latin-hypercube designs for computer experiments

Sliced Orthogonal Array-Based Latin Hypercube Designs

Construction of Optimal Mixed-Level Uniform Designs

Asymptotically optimal maximin distance Latin hypercube designs

Construction of Marginally Coupled Designs by Subspace Theory

Search for Minimum Aberration Designs with Uniformity

Sliced Latin hypercube designs with arbitrary run sizes

Expected $L_2-$discrepancy bound for a class of new stratified sampling models

Construction of uniform U-designs

Bi-directional Sliced Latin Hypercube Designs

The generalization of Latin hypercube sampling

Is a Transformed Low Discrepancy Design Also Low Discrepancy?

Optimizing Latin hypercube design for sequential sampling of computer experiments

Transforming the Challenge of Constructing Low-Discrepancy Point Sets into a Permutation Selection Problem

SF-SFD: Stochastic Optimization of Fourier Coefficients to Generate Space-Filling Designs

Flexible sliced designs for computer experiments

Construction of uniform designs without replications.

Constructing Optimal $L_{\infty}$ Star Discrepancy Sets

Sampling in CMA-ES: Low Numbers of Low Discrepancy Points