|Title:||Minimum breakdown designs in blocks of size two||Authors:||SHIN-FU TSAI||Keywords:||Block design | Connectedness | Missing value | Robust design | Two-color microarray||Issue Date:||2013||Publisher:||ELSEVIER SCIENCE BV||Journal Volume:||143||Journal Issue:||1||Start page/Pages:||202||Source:||Journal of Statistical Planning and Inference||Abstract:||
In scientific investigations, there are many situations where each two experimental units have to be grouped into a block of size two. For planning such experiments, the variance-based optimality criteria like A-, D- and E-criterion are typically employed to choose efficient designs, if the estimation efficiency of treatment contrasts is primarily concerned. Alternatively, if there are observations which tend to become lost during the experimental period, the robustness criteria against the unavailability of data should be strongly recommended for selecting the planning scheme. In this study, a new criterion, called minimum breakdown criterion, is proposed to quantify the robustness of designs in blocks of size two. Based on the proposed criterion, a new class of robust designs, called minimum breakdown designs, is defined. When various numbers of blocks are missing, the minimum breakdown designs provide the highest probabilities that all the treatment contrasts are estimable. An exhaustive search procedure is proposed to generate such designs. In addition, two classes of uniformly minimum breakdown designs are theoretically verified. © 2012 Elsevier B.V.
|Appears in Collections:||農藝學系|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.