|
Department of Biostatistics Colloquium Series 2010 - 2011 |
ABSTRACT: In recent years, there has been a revolution in biotechnology with clinicians, medical researchers and basic scientists now having access to an amazing array of tools. For example, for an individual patient, we can obtain gene sequence information at over a million locations, measure tens of thousands of time-varying gene expression levels, assay levels of exposure to numerous environmental contaminants, monitor heart and brain functioning using high-resolution imaging, and record countless other types of massive-dimensional data. This information revolution is extremely exciting and has already led to fundamental breakthroughs in understanding of complex diseases, such as cancer and diabetes. The available data also has great potential in developing targeted therapies for dramatically improving disease prevention and treatment. However, there is a daunting problem in realizing this potential - namely, finding the informative needles in the massive-dimensional haystack of data we collect. In this talk, I describe how we can build automated Bayesian methods for scientific and predictive learning from high-dimensional and complex data collected on few individuals. A variety of applications are described for predicting disease and adverse health outcomes based on images, curves and high-dimensional biomarkers collected over time.
ABSTRACT: The rapidly growing national interest in comparative effectiveness research (CER) hold the promise of providing vital new information for physicians, patients, and policymakers about the relative efficacy, safety, and cost-effectiveness of alternative treatments for common conditions. A $1.1 billion infusion of support from the 2009 stimulus bill, along with up to $500 million per year in funding for the just-formed Patient-Centered Outcomes Research Institute, will fuel an ambitious new investigative agenda in this area, and potentially fill important voids in our knowledge of optimal patient care. Several important issues are emerging as this new field takes shape, which will be addressed in this presentation. What are the most appropriate comparators, and how are they to be chosen? How are superiority or non-inferiority best quantified? What role should be played by observational studies vs. randomized trials in addressing this agenda? Can CER adequately explore different responses of various subgroups that are defined demographically, clinically, or genetically? Will faceless Washington bureaucrats use this research to convene death panels to deny care to peoples’ grandmothers? This presentation will provide a firstinning assessment of where these questions stand from the perspective of one participant in this research.
ABSTRACT: The concept of a community is central to social network analysis, and thus a large body of work has been devoted to identifying community structure. For example, a community may be thought of as a set of WebPages on related topics, a set of advertisers in a similar economic market, or more generally as a set of nodes in a network more similar amongst themselves than with the remainder of the network. Motivated by difficulties we experienced at actually finding meaningful communities in very large real-world networks, we have performed a large scale analysis of a wide range of social and information networks. Our main methodology used local spectral methods and involved computing isoperimetric properties of the networks at various size scales---a novel application of ideas from statistics and scientific computation to internet data analysis, which required the development of new algorithmic tools, as well as the reinterpretation of the statistical basis underlying traditional spectral approximation algorithms.
Our empirical results suggest a significantly more refined picture of community structure than has been appreciated previously. Our most striking finding is that in nearly every network dataset we examined, we observe tight but almost trivial communities at very small size scales, and at larger size scales, the best possible communities gradually "blend in" with the rest of the network and thus become less "community-like." This behavior is not explained, even at a qualitative level, by any of the commonly-used network generation models. Moreover, this behavior is exactly the opposite of what one would expect based on experience with and intuition from expander graphs, from graphs that are well-embeddable in a low-dimensional structure, and from small social networks that have served as test beds of community detection algorithms. Possible mechanisms for reproducing our empirical observations will be discussed, as will implications of these findings for clustering, classification, and more general data analysis in modern large social and information networks.
ABSTRACT: In addition to may problems of policy, financing, staffing and parochialism of the US healthcare system, we also suffer from poor integrative technologies that could improve health care quality and equity. The Federal government is making a big bet that the adoption of electronic health records will significantly help with all these problems, but an NRC study published in 2009 suggests caution in these expectations. I will review some of the findings of that study, and describe current as well as additionally needed research in computer science and its applications that can make a big difference in achieving the goals of a truly safe, effective, patient-centered, timely, efficient and equitable health care system.
ABSTRACT: None Given
ABSTRACT: None Given
ABSTRACT: None Given
ABSTRACT: None Given
|
Click
here for past schedules |
Biostatistics
Webmaster |