 2011 Annual Meeting Date: Friday, April 29, 2011, 12:30-5:00 p.m. Location: Santa Fe Hilton, 100 Sandoval Street, Santa Fe, NM, Chapel/Aspen Room Cost: $40 (discounted cost is $20 for full-time students). Pre-registration To register, e-mail Fletcher Christensen at kurisu@stat.unm.edu to let him know you plan to attend. The registration fee is payable at the meeting. Please register early to ensure that we have seats and refreshments for everyone. Note: we will also accept additional voluntary contributions to help pay for science fair awards and a future short course. Program 12:30-1:00 Registration, Santa Fe Hilton, Chapel/Aspen Room 1:00-1:10 Welcome, Oleg Makhnin 1:10-2:00 Talks 1 and 2 Jim Wendelberger, Urban Science: Statistical Engineering as Practiced at Urban Science Dave Collins, Statistical Sciences Group, Los Alamos National Laboratory: Reliability Modeling of Nuclear Power Plant Subsystems Using Statistical Flowgraphs 2:00-2:20 Break 2:20-3:10 Talks 3 and 4 Bruce Layman, Design Agency Quality Engineering Group, Los Alamos National Laboratory: Tails of the Poisson Erik B. Erhardt, The Mind Research Network: A Bayesian Framework for Stable Isotope Mixing Models 3:10-3:30 Break 3:30-4:20 Talks 5 and 6 Sham Bhat, Statistical Sciences Group, Los Alamos National Laboratory: Bayesian Inference for Complex Computer Models and Large Multivariate Spatial Data for Climate Science Oleg Makhnin, Department of Mathematics, New Mexico Tech: Multisite precipitation generator based on switching weather states 4:20-4:30 Break 4:30-5:00 Presentation of award, business meeting Abstracts Statistical Engineering as Practiced at Urban Science Jim Wendelberger, Statistical Analysis, Urban Science Several statistical engineering solutions are described. Included in this description are the engineering constraints under which a statistical engineering solution was arrived at. These include the many considerations that are relevant in statistical engineering, such as, Quantitative Theory, Technology, Management System, Statistical Tools, Legal Aspects, Political Aspects, Software Constraints, Data Availability, Cost (time, money, political, etc.), Computational (Memory, speed, storage), End Result (Report, PowerPoint, Verbal, Software, etc.), Model Constraints (External restrictions), Model Assumptions (External Tenability), Client Constraints (May affect any of the above and possibly more), Delivery Vehicle and Deliverables. Reliability Modeling of Nuclear Power Plant Subsystems Using Statistical Flowgraphs Dave Collins, Statistical Sciences Group, Los Alamos National Laboratory Nuclear power plants (NPPs) play an important role in energy security and reduction of airborne pollutants. As part of the U.S. Department of Energy’s reactor sustainability program, we are developing models to characterize the reliability and safety of NPP piping subsystems. Subsystems are represented as statistical flowgraphs, with vertices representing states of partial or complete failure and edges representing probability distributions for transitions between states. Failure transitions are driven by processes based on material properties of the pipes, and the physical and chemical dynamics of the fluid being carried. Repair transitions are based on detection of leakage by visual inspection, or non-visible flaws by radiography or ultrasound. Given the complexity of the transition processes, we model the subsystems as semi-Markov processes, using the flowgraph methodology to solve for quantities of interest such as the hazard rate for pipe rupture. Tails of the Poisson Bruce Layman, Design Agency Quality Engineering Group, Los Alamos National Laboratory The infinite sums involved in calculating the tails of the cumulative Poisson distributions can be calculated with extreme accuracy using either the Type I or Type II Rational Approximations developed in this paper. These approximations follow elegant patterns that make their calculations extremely fast and accurate. Recursion formula connecting the various orders of these approximations can used to increase accuracy. Correction formula for the Rational Approximations can also be used to increase accuracy. The techniques used to find the Rational Approximations for the Poisson distributions are quite general and can be used to approximate a wide variety of infinite sums including the Error Function. A Bayesian Framework for Stable Isotope Mixing Models Erik B. Erhardt, The Mind Research Network Stable isotope sourcing is used to estimate proportional contributions of sources to a mixture, such as in the analysis of animal diets and plant nutrient use. Statistical methods for inference on the diet proportions using stable isotopes have focused on the linear mixing model. Existing frequentist methods provide inferences when the diet proportion vector can be uniquely solved for in terms of the isotope ratios. Bayesian methods apply for arbitrary numbers of isotopes and diet sources but existing models are somewhat limited as they assume that trophic fractionation or discrimination are estimated without error or that isotope ratios are uncorrelated. We present a Bayesian model for the estimation of mean diet that accounts for uncertainty in source means and discrimination and allows correlated isotope ratios. This model is easily extended to allow the diet proportion vector to depend on covariates, such as time. Two examples are used to illustrate the methodology. Bayesian Inference for Complex Computer Models and Large Multivariate Spatial Data for Climate Science Sham Bhat, Statistical Sciences Group, Los Alamos National Laboratory Computer model calibration involves combining information from a complex computer model with physical observations of the process. Computer model output is often in the form of multiple spatial fields, particularly in climate science. We study an effective inferential approach by using Gaussian processes to emulate the computer model, linking the calibration parameters with the multivariate spatial observations. Then we infer the calibration parameters using Bayesian methods, while incorporating more flexible approaches allowing for non-linear relationships among spatial fields and non-separable covariance functions. In addition, we incorporate the uncertainty due to model discrepancy and measurement error into our inference and predictions, which usually results in more accurate and sharper inference of the calibration parameter and improved characterization of uncertainties. We utilize kernel mixing and matrix identities in order to make computations tractable for large spatial data sets. We apply our approach to infer vertical diffusivity, a climate model parameter from which we obtain projections of the Atlantic Meridional Overturning Circulation (MOC). Multisite precipitation generator based on switching weather states Oleg Makhnin, Department of Mathematics, New Mexico Tech In a recent paper, Makhnin and McAlister described a multisite precipitation generator employing multivariate autoregression and truncated and power-transformed (TPT) Normal distribution of the precipitation values. The statistics for the data produced by the generator match the historically observed data fairly well. However, the distribution of dry spells produced by this generator is still not completely satisfying. Here, an extension of the above model is discussed where the current region-wide "weather state" (e.g. "dry" or "wet") is produced according to a Markov switching model, and the multivariate autoregressions depend on the current weather state. |