A Note on the Choice and the Estimation of Kriging Models for the Analysis of Computer Experiments

David Ginsbourger (1) , Delphine Dupuy (1) , Anca Badea (1) , Laurent Carraro (1) , Olivier Roustant (1)

1 3MI-ENSMSE - Département Méthodes et Modèles Mathématiques pour l'Industrie

Our goal in the present work is to give an insight on some important questions to be asked when choosing a kriging model for the analysis of numerical experiments. We are especially concerned about the cases where the size of the design of experiments is small relatively to the algebraic dimension of the inputs. We ¯rst ¯x the notations and recall some basic properties of kriging. Then we expose two experimental studies on subjects that are often skipped in the ¯eld of com- puter simulation analysis: the lack of reliability of likelihood maximization with few data, and the consequences of a trend misspeci¯cation. We ¯nally propose a case study, with the application of an original kriging method where an additive model is used as external trend.

Mots clés: D.A.C.E

Mots clés: Kriging; External Drift; Likelihood Methods; D.A.C.E.

David Ginsbourger, Delphine Dupuy, Anca Badea, Laurent Carraro, Olivier Roustant. A Note on the Choice and the Estimation of Kriging Models for the Analysis of Computer Experiments. Joint ENBIS-DEINDE 2007 conference Computer experiments versus physical experiments, Apr 2007, Turin, Italy. ⟨hal-00409749⟩