Multistage design procedures for identifying two-factor interactions, when higher effects are negligible (Q1300929)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Multistage design procedures for identifying two-factor interactions, when higher effects are negligible |
scientific article |
Statements
Multistage design procedures for identifying two-factor interactions, when higher effects are negligible (English)
0 references
15 July 2001
0 references
The authors consider experiments in which \(t \geq 2\) test treatments are to be compared to a set of \(s \geq 2\) standard treatments (\(s \leq t\)) using a connected block design with \(b\) blocks of size \(k\) each. The labels \(1,\ldots,s\) are used for the standard treatments and \(s+1,\ldots,v,\) (\(v=s+t\)) are used for the test treatments. The observations are assumed to follow the usual additive model with uncorrelated errors having a common variance \(\sigma^2\), i.e., \[ Y_{ijp} = \alpha_i + \beta_j + E_{ijp}, \] where \(\alpha_i\) represent the treatment effects and \(\beta_j\) the block effects. Let \(D(s,t;b,k)\) denote the class of all connected block designs with \(v=s+t\) treatments arranged in \(b\) blocks of size \(k\) each. Further, let \(D(s,t;b,k;n_0)\) denote the subclass of \(D(s,t;b,k)\) with a total of \(n_0\) experimental units assigned to the \(s\) standard treatments. A design \(d^*\) is said to be MV-optimal in a class of designs \(D\) for comparing the test treatments to the standard treatments, if for any other design \(d \in D\), \[ \max \text{Var}_{d^*}(\hat \alpha_i - \hat \alpha_j) \leq \max \text{Var}_d(\hat \alpha_i - \hat \alpha_j),\quad 1 \leq i \leq s, \;1 \leq j \leq s. \] The authors establish a set of sufficient conditions for a design to be MV-optimal in the classes \(D(s,t;b,k)\) and \(D(s,t;b,k;n_0)\), and show, via examples, the use of these sufficient conditions to demonstrate the optimality of certain block designs which are orthogonal or nearly-orthogonal for estimating the differences between the test and standard treatments.
0 references
search designs
0 references
model selection
0 references