Abstract
We introduce a new class of designs, marginally coupled designs, for computer experiments with both qualitative and quantitative variables. These designs maintain an economic run size with attractive space-filling properties. The design points for quantitative factors form a Latin hypercube design. In addition, for each level of any qualitative factor of a marginally coupled design, the corresponding design points for quantitative factors form a small Latin hypercube design. Existence of the proposed designs is studied. Constructions are provided for various types of designs with qualitative factors.
Original language | English (US) |
---|---|
Pages (from-to) | 1567-1581 |
Number of pages | 15 |
Journal | Statistica Sinica |
Volume | 25 |
Issue number | 4 |
DOIs | |
State | Published - Oct 2015 |
All Science Journal Classification (ASJC) codes
- Statistics and Probability
- Statistics, Probability and Uncertainty
Keywords
- Difference scheme
- Fold-over design
- Latin square
- Mixed level
- Orthogonal array
- Resolvable orthogonal array