Abstracts for Keynote Sessions
Lawrence H. Cox
National Center for Health Statistics
Tables are a staple of statistical analysis. Several problems related to statistical analysis in tables may be modeled in the language of mathematical programming, including the following. (1) Rounding some or all table entries to a fixed integer rounding base. An important special case in statistics is rounding noninteger entries to integers, as when, given a table of noninteger MLEs for entries in a contingency table, an unbiased integer-valued estimate is sought, e.g., to achieve a two-way stratified random sample in survey sampling. (2) Achieving confidentiality protection in tables of count or aggregate magnitude data, as when an appropriate rounding, perturbation, suppression or imputation of a table or a confidentiality audit on a statistical data base query system is sought. (3) Generating a random sample of contingency tables subject to fixed marginal totals, e.g., using MCMC, as when a test of hypothesis for a specific table is sought. While modeling of these problems is straightforward, and informative, results based on mathematical programming theory demonstrate that in general these models are expected to be extremely difficult to compute (NP-hard) and in some cases may fail to have a solution. An exception is the mathematical network, which represents a class of tables including two-way tables. In this talk, we discuss the statistical problems described above, their corresponding mathematical and network models, and issues affecting their computability or insolubility.
Eugene M. Isenberg School of Management
University of Massachusetts Amherst
Industrial statistics is at a cross road. Now that manufacturing employ fewer than 10% of the US workforce, and much traditional manufacturing is moving off shore, industrial statisticians need to rethink their role. Trends in computing, statistical software and the internet have also changed how statistics is applied in business and industry. In this talk I will provide an overview of how the role of statisticians working in industry is changing and what the challenges are to the profession. I will discuss changes to what we traditionally consider the domain of quality control and quality management and what role statistics play in the knowledge economy. Illustrations from business and industry will be provided.