What is it about?
Recently multi-objective and compound design criteria have been proposed to help the design of multi-factor experiments focusing on several objectives. Often, one of the objectives is the prediction of the response under different settings of the experimental factors. In this paper, it is argued that predictions of differences in responses might be more relevant. A compound design criteria, that involves parameter estimation, predictions of responses and differences in responses, is proposed. The criteria allow the focus on point estimation/prediction as well as interval estimation/prediction. The paper also proposes variations of the existing graphical tools, such as variance dispersion and fraction design space graphs, to better depict the prediction performances of the designs being compared. Two examples illustrate the methods.
Photo by Clark Van Der Beken on Unsplash
Why is it important?
Experiments provide important information for discoveries in many research areas. Careful planning of an experiment is very important in order to obtain informative answers to the research questions. As the researcher, usually, would like to answer more than one question with one experiment, a design with several good properties is required.
Read the Original
This page is a summary of: Compound optimality criteria and graphical tools for designs for prediction, Quality and Reliability Engineering International, June 2022, Wiley, DOI: 10.1002/qre.3150.
You can read the full text:
The following have contributed to this page