Brown University Center for Statistical Sciences Seminar
Candidate for Assistant Professor (Research) Position at the Department of Community Health and the Center for Statistical Sciences. | |
Abstract: One of the important issues of statistical analysis is to compute accurate estimates for parameters of interests. Confidence intervals are preferred than plain point estimates. Classical confidence intervals are constructed based on large sample normality of the estimates. This will fail when there are only small samples and the parameter of interest is close to boundary values. ABC (Approximate Bootstrap Confidence) Interval is a very general theory to compute higher-order C.I.s for small sample sizes and arbitrary smooth parameter of interests. However, due to the complexity of the mathematics of this theory, it takes considerable mathematical sophistication to visualize and formulate the statistical problem at hand into the framework of ABC theory.
In this talk, we will present the basic concepts and algorithms of the ABC theory. Then we will show how to apply ABC theory to compute accurate confidence interval for the classical Stress-Strength model under Gamma distribution, particularly how to formulate the problem into ABC framework.
Seminar on Nonlinear Waves
Lefschetz Center for Dynamical Systems Seminar
Abstract: One of the emerging technical issues in coastal oceanography is how to combine disparate data sources with model output to produce reliable nowcasts of the surface flow. I describe a spectral method, called normal mode analysis, which is well suited for this task. The method is also useful for time and space filtering of data fields. The method is described and its attributes are illustrated with a discussion of nowcasts of the surface velocity in Monterey Bay for 1-9 August 1994. The nowcasts were obtained by blending HF radar observations with output from a primitive equation model.
Brown Analysis Seminar
Brown University Center for Statistical Sciences Seminar
Quality of Life Measures for Evaluating Diagnostic Imaging | |
Abstract:
Advances in diagnostic imaging can affect patients' quality of life on several scales. New imaging techniques may involve less (or more) patient discomfort and inconvenience on short time scales associated with the acquistion of the image itself. Improved diagnostic information may alter treatment selection, changing patients quality of life associated with treatment and its side effects on moderate time scales of days to months. And finally, impoved diagnostic information may lead to better long term quality of life outcomes on the order of years. Measurement of quality of life outcomes associated with diagnostic imaging particularly on short time scales is a nascent research area. In this talk we will review some measurement approaches to quantify these effects on quality of life for cost-effectiveness evaluation of new imaging techniques.
Spnsored by: The Bruce M. Bigelow Class of 1955 Lecture Series, The Charles K. Colver Lectureships and Publication Fund and the Department of Diagnostic Imaging at Brown University and Rhode Island Hospital/Lifespan
Brown University Center for Statistical Sciences Seminar
Quality of Life Measures for Evaluating Diagnostic Imaging | |
Abstract: Various screening tests for ascertaining characteristics or conditions have been in widespread use, especially in clinical practice or medical research. Concerns with the quality and the costs of the tests, especially for mass screening, raised a need to administer them in a most efficient manner.
e consider the case where the accuracy of a screening test depends on the values of the test outcome variables, selected to classify the screening outcome variables, selected to classify the screening outcome as being indicative of presence or absence of the characteristic to be detected. Following the parametric Bayesian decision-theoretic approach proposed by Geisser (1998), the problem of optimal dichotomization for screening tests is considered. For situations where the dichotomization is influenced by certain covariates subject to random errors, we also suggest optimal methods based on the proposed approach. Monte Carlo optimization theory is applied to cope with the difficulty in searching for optimal dichotomizers.
Scientific Computing Seminar
PDE Seminar
Abstract: The classical Chapman-Enskog expansion provides a bridge between the kinetic description of gas dynamics and continuum mechanics. However truncation of this expansion beyond (first) Navier Stokes order yields instability of the rest state. A special case of Chapman-Enskog is obtained when the constraint of constant density is imposed: the Rivlin-Ericksen fluid which of course exhibits the same instability phenomena upon truncation. This talk illustrates how an approximate sum of the Chapman-Enskog expansion eliminates the instability paradox via derivation of a generalized Clausius-Duhem inequality.
Department of Mathematics Colloquium
<--- 1999 Index