Efficient and Robust Designs Based on Ordered Units

University Of Illinois At Chicago, Chicago IL

Investigators

Abstract

Abstract DMS-0204532 PI: Dibyen Majumdar Title: Efficient and Robust Designs Based on Ordered Units The goal of this project is to find efficient designs for experiments conducted over time to compare different treatments, varieties of objects or methods. A substantial part of the project deals with crossover experiments, where it often happens that subjects drop out prior to the conclusion of the experiment, thereby reducing the information content of the data, or even jeopardizing the validity of the conclusions. Here the aim is to identify efficient designs that are largely robust to subject dropout. Another part of the project deals with experiments where a trend over time must be accounted for in the model and analysis. Here the aim is to identify efficient designs that are flexible enough to accommodate a wide variety of experimental conditions and models that arise in real applications. Proper statistical design is a critical aspect of scientific experiments; it ensures that the conclusions, based on observed data, remain valid in the larger context. This project deals with experiments in a variety of fields, including biomedical, agricultural and industrial experiments, that are conducted over time. As the experiment progresses, uncontrollable factors could influence the outcome. The aim of this project is to develop designs that perform well in diverse experimental conditions as well as designs that are robust against the inadvertent loss of experimental subjects before the conclusion of the experiment.