Zhang, Qiong; Qian, Peter Z. G. Designs for crossvalidating approximation models. (English) Zbl 1279.62158 Biometrika 100, No. 4, 997-1004 (2013). Summary: Multifold crossvalidation is routinely used for assessing the prediction error of an approximation model for a black-box function. Despite its popularity, this method is known to have high variability. To mitigate this drawback, we propose an experimental design approach that borrows Latin hypercube designs to construct a structured crossvalidation sample such that the input values in each fold achieve uniformity. Theoretical results show that the estimate of the prediction error of the proposed method has significantly smaller variability than its counterpart under independent and identically distributed sampling. Numerical examples corroborate the theoretical results. Cited in 3 Documents MSC: 62K10 Statistical block designs 05B15 Orthogonal arrays, Latin squares, Room squares 65C60 Computational problems in statistics (MSC2010) Keywords:black-box functions; computer experiments; design of experiments; Latin hypercube designs; meta-models; sliced Latin hypercube designs; surrogate models PDFBibTeX XMLCite \textit{Q. Zhang} and \textit{P. Z. G. Qian}, Biometrika 100, No. 4, 997--1004 (2013; Zbl 1279.62158) Full Text: DOI Link