Robust Uncertainty Quantification and Scalable Computation for Computer Models with Massive Output

Uncertainty quantification (UQ) is both an old and new concept. The current novelty lies in the interactions and synthesis of mathematical models, computer experiments, statistics, field/real experiments, and probability theory, with a particular emphasize on the large-scale simulations by computer models. The challenges not only come from the complication of scientific questions, but also from the size of the information. It is the focus in this thesis to provide statistical models that are scalable to massive data produced in computer experiments and real experiments, through fast and robust statistical inference.

Chapter 2 provides a practical approach for simultaneously emulating/approximating massive number of functions, with the application on hazard quantification of Soufri\`{e}re Hills volcano in Montserrate island. Chapter 3 discusses another problem with massive data, in which the number of observations of a function is large. An exact algorithm that is linear in time is developed for the problem of interpolation of Methylation levels. Chapter 4 and Chapter 5 are both about the robust inference of the models. Chapter 4 provides a new criteria robustness parameter estimation criteria and several ways of inference have been shown to satisfy such criteria. Chapter 5 develops a new prior that satisfies some more criteria and is thus proposed to use in practice.

WriteSim TCExam--an open source text simulation environment for training novice researchers in scientific writing.

BACKGROUND: The ability to write clearly and effectively is of central importance to the scientific enterprise. Encouraged by the success of simulation environments in other biomedical sciences, we developed WriteSim TCExam, an open-source, Web-based, textual simulation environment for teaching effective writing techniques to novice researchers. We shortlisted and modified an existing open source application - TCExam to serve as a textual simulation environment. After testing usability internally in our team, we conducted formal field usability studies with novice researchers. These were followed by formal surveys with researchers fitting the role of administrators and users (novice researchers) RESULTS: The development process was guided by feedback from usability tests within our research team. Online surveys and formal studies, involving members of the Research on Research group and selected novice researchers, show that the application is user-friendly. Additionally it has been used to train 25 novice researchers in scientific writing to date and has generated encouraging results. CONCLUSION: WriteSim TCExam is the first Web-based, open-source textual simulation environment designed to complement traditional scientific writing instruction. While initial reviews by students and educators have been positive, a formal study is needed to measure its benefits in comparison to standard instructional methods.

Scientific writing: a randomized controlled trial comparing standard and on-line instruction.

BACKGROUND: Writing plays a central role in the communication of scientific ideas and is therefore a key aspect in researcher education, ultimately determining the success and long-term sustainability of their careers. Despite the growing popularity of e-learning, we are not aware of any existing study comparing on-line vs. traditional classroom-based methods for teaching scientific writing. METHODS: Forty eight participants from a medical, nursing and physiotherapy background from US and Brazil were randomly assigned to two groups (n = 24 per group): An on-line writing workshop group (on-line group), in which participants used virtual communication, google docs and standard writing templates, and a standard writing guidance training (standard group) where participants received standard instruction without the aid of virtual communication and writing templates. Two outcomes, manuscript quality was assessed using the scores obtained in Six subgroup analysis scale as the primary outcome measure, and satisfaction scores with Likert scale were evaluated. To control for observer variability, inter-observer reliability was assessed using Fleiss's kappa. A post-hoc analysis comparing rates of communication between mentors and participants was performed. Nonparametric tests were used to assess intervention efficacy. RESULTS: Excellent inter-observer reliability among three reviewers was found, with an Intraclass Correlation Coefficient (ICC) agreement = 0.931882 and ICC consistency = 0.932485. On-line group had better overall manuscript quality (p = 0.0017, SSQSavg score 75.3 +/- 14.21, ranging from 37 to 94) compared to the standard group (47.27 +/- 14.64, ranging from 20 to 72). Participant satisfaction was higher in the on-line group (4.3 +/- 0.73) compared to the standard group (3.09 +/- 1.11) (p = 0.001). The standard group also had fewer communication events compared to the on-line group (0.91 +/- 0.81 vs. 2.05 +/- 1.23; p = 0.0219). CONCLUSION: Our protocol for on-line scientific writing instruction is better than standard face-to-face instruction in terms of writing quality and student satisfaction. Future studies should evaluate the protocol efficacy in larger longitudinal cohorts involving participants from different languages.

BAYESIAN MODEL SEARCH AND MULTILEVEL INFERENCE FOR SNP ASSOCIATION STUDIES.

Technological advances in genotyping have given rise to hypothesis-based association studies of increasing scope. As a result, the scientific hypotheses addressed by these studies have become more complex and more difficult to address using existing analytic methodologies. Obstacles to analysis include inference in the face of multiple comparisons, complications arising from correlations among the SNPs (single nucleotide polymorphisms), choice of their genetic parametrization and missing data. In this paper we present an efficient Bayesian model search strategy that searches over the space of genetic markers and their genetic parametrization. The resulting method for Multilevel Inference of SNP Associations, MISA, allows computation of multilevel posterior probabilities and Bayes factors at the global, gene and SNP level, with the prior distribution on SNP inclusion in the model providing an intrinsic multiplicity correction. We use simulated data sets to characterize MISA's statistical power, and show that MISA has higher power to detect association than standard procedures. Using data from the North Carolina Ovarian Cancer Study (NCOCS), MISA identifies variants that were not identified by standard methods and have been externally "validated" in independent studies. We examine sensitivity of the NCOCS results to prior choice and method for imputing missing data. MISA is available in an R package on CRAN.