These papers focuss on applied interests. They may be partitioned into three parts: design strategies in fractional factorial experiments, the additive variety-block model, and modelling work on Parkinsonian data.
Summary. It is often known in advance that certain subsets of factors act independently upon a response. Such information can be used to estimational advantage by aliasing low order effects with such zero interactions. We find the best $2^{n-k}$ fractions for the case when the factors can be partitioned into two classes such that non-zero interactions may exist only between classes but not within a class.
Summary. When studying a response as a function of several factors, engineering reasons or other deterministic considerations often imply that interactions between certain factors do not exist. This prompts advantageous exploitation of the aliasing pattern in fractional factorial designs. The general case when factors (each at two levels) can be partitioned into two classes, with no interactions present between the two classes, is treated in this paper.
Summary. In practical experiments researchers usually wish to control both the cost of the experiment and the precision of the parameters to be estimated. When the average cost per observation is known, the overall cost becomes proportional to the number of observations. This article offers theory and algorithms that take these experimental concerns as input data and yield as output the actual design which conforms (as much as can be expected) to the specified requirements. This accomplishment is restricted to the additive model of block designs.
Center of pressure electronic platform testing is proposed as an affordable early diagnostic tool for persons at risk of Parkinson's disease. Stiffness measures based on statistical concepts are used in such a diagnosis.
A constructive basis is found for the subspace of unbiased estimates of $<\alpha_i-\alpha_j>$ in the additive variety-block model. Graph-theoretic and geometric interpretations are then given to the best linear unbiased estimates of elementary variety constrasts. The paper offers an understanding of how new statistical information is incorporated by way of this model directly in terms of the underlying graph structure.