Author

Ying Pei

Date of Graduation

2015

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Industrial and Managements Systems Engineering

Committee Chair

Yang Feng

Committee Co-Chair

Wafik Iskander

Committee Member

Majid Jaridi

Abstract

One of the major challenges with toxicology studies of nanomaterials (NMs), compared to traditional materials or chemicals, lies in the large NM variety (or sources) caused by their various physico-chemical properties. How to efficiently design multi-source biological experiments for the toxicity characterization of NMs in terms of their exposure-response profiles? This work intends to address this question by a two-stage experimental design procedure, which is developed based on the statistical model, stochastic kriging with qualitative factors (SKQ). With a given experimental budget, the SKQ-based design method aims at achieving the highest-quality SKQ, which synergistically models the exposure-response data from multiple sources (e.g., NM types). The method determines the experimental design (that is, the sampling location as well as allocation) in such a way that the resulting sampling data allow SKQ to realize its maximum potential to pool information across multiple sources for efficient modeling. Built in a two-stage framework, which enables a learning process of the target exposure-response relationships, the SKQ-based design procedure also inherits the general advantages of stochastic kriging in the sense that the design is particularly tailored to model the possibly nonlinear and complex relationships and heterogeneous data variances. Through simulation studies, the efficiency of the SKQ-based procedure for multi-source experiments is demonstrated over the two alternative design methods.

Share

COinS