Prediction capabilities of optimum designs: completely randomized and split-plot designs

• Grant number: 13/09282-9
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: August 01, 2013
• End date: September 30, 2014
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Applied Probability and Statistics
• Principal Investigator: Luzia Aparecida Trinca
• Grantee: César Buchile Abud de Oliveira
• Host Institution: Instituto de Biociências (IBB). Universidade Estadual Paulista (UNESP). Campus de Botucatu. Botucatu , SP, Brazil

Abstract

The popularity and applicability of optimum designs have reached considerable growth in recent years. The use of an optimum or efficient design is very important since it means the maximum of information given the fixed resources will be attained. Designs tailored for the practical problem are now a reality given compound criteria suggest by Gilmour and Trinca (2012). Other desired properties not yet included in compound criteria are the error margin of response predictions. In this project, we propose to study the prediction properties of optimum designs. Such properties can be studied through variance dispersion graphs in the experimental region, the so-called VDG, and DVDG, the last one referring to the variance of differences between estimated responses. We propose to extend the methods to calculate VDG e DVDG to designs in two strata (split-plot). To construct the graphs the maximum, minimum, and average variances of quantities of interest are calculated considering hyper-spheres of increasing radius throughout the experimental region. C and R languages will be combined in order the calculate the required quantities. (AU)