Joint Research Conference
June 24-26, 2014
Factorial designs are often used in many scientific investigations when the interest of experimenters lies in the study of effects of two or more factors simultaneously. Fractional factorial designs, which consist of a subset of runs in full factorial designs, are commonly used in practice for the economic reasons. Effect aliasing is a consequence of using fractional factorial designs. Effect aliasing for the fixed effects model, which is usually used to model the data from a physical experiment, has been extensively studied in the literature and well understood. An alternative model, called the Gaussian random field model, for the data is the stochastic process approach appeared in some Bayesian design literature. The impact of effect aliasing for the Gaussian random field model has not received adequate attention in the literature. Part of the reason for this is that the Gaussian random field model is usually not characterized by a linear model structure that is inherent in the fixed effects model. In this work, we establish a kind of linear model structure for the Gaussian random field model and discuss effect aliasing in the Gaussian random field model for p-level regular fract
Key words: Fractional factorials, Bayesian design, computer experiments