ENBIS-12 in Ljubljana

9 – 13 September 2012 Abstract submission: 15 January – 10 May 2012

My abstracts

 

The following abstracts have been accepted for this event:

  • Implementing Design and Analysis of Experiments in the U.S. Department of Defense Testing Community

    Authors: Laura Freeman (The Institute of Defense Analyses)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Design of experiments, acquisition, statistical challenges, Department of Defense
    Submitted at 27-May-2012 23:17 by Laura Freeman
    Accepted
    10-Sep-2012 12:50 Implementing Design and Analysis of Experiments in the U.S. Department of Defense Testing Community
    The Department of Defense (DoD) acquires and tests some of the world’s most complex and costly systems. However, until recently, there has been no standard methodology applied to testing military systems. In 2010, the Director, Operational Test and Evaluation (DOT&E) issued guidance on the use of Design of Experiments (DOE) in operational testing. DOT&E is responsible for the oversight of all operational tests for major DoD acquisition systems. Additionally, DOT&E is responsible for reporting on the outcome of these tests to the Secretary of Defense and Congress.
    This presentation provides an overview of the design and analysis of experiments within the US DoD acquisition process. It discusses recent initiatives to increase the scientific rigor of operational tests plans using design and analysis of experiments. The presentation covers several examples of test plans constructed using DOE on a variety of systems. Finally, it highlights the cultural, educational, and statistical challenges of implementing more statistical rigor in a large and complex organization like the DoD test and evaluation community.
  • Issues in Planning Experiments for Highly Constrained Regions

    Authors: Geoff Vining (Virginia Tech)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: constraints, design of experiments, optimal design, experimental protocol
    Submitted at 27-May-2012 23:20 by Geoff Vining
    Accepted
    10-Sep-2012 11:50 Issues in Planning Experiments for Highly Constrained Regions
    Engineers and scientists often encounter experimental situations involving constraints on the factor levels and/or the total number of experimental runs, in some cases, rather extreme constraints. Using such classical experimental design strategies as 2^k, 2^(k-p), or a central composite designs appear rather naïve, especially in light of the number of statistical software packages that generate “optimal’ designs. Unfortunately, use of such optimal designs often is equally as naïve.

    This talk outlines the basic issues that experimenters must address as they plan experiments involving constrained design spaces. Such issues include the concept and impact of sequential experimentation, the distinction between the “region of operability” and the actual “experimental region,” the specific nature of the experimental region, the impact of the presumed model to be fit, issues of model robustness, and why “optimal” designs are never truly optimal.

    This talk outlines an appropriate thought process for determining appropriate experimental designs for constrained regions. The goal is to guide practitioners to make intelligent decisions appropriate for determining experimental protocols.
  • Lies, Damn Lies and Business Statistics! What can NSIs do to Reveal the Truth?

    Authors: Steve MacFeely (Central Statistics Office)
    Primary area of focus / application: Mining
    Keywords: information, stories, analysis
    Submitted at 28-May-2012 17:39 by Steve MacFeely
    Accepted
    Official business statistics, like all official statistics, are inadvertently or deliberately abused on a routine basis. While one can never determine with absolute certainty the reason, it appears that most abuses occur mistakenly, because users did not fully understand the data, the associated metadata or the underlying concepts.
    This paper argues that, in a time of abundant data, National Statistical Institutes (NSIs) must rethink their mission and in particular their communication and analysis strategies. Fundamental to this reassessment is a simple question; is the role of the NSI simply to provide independent, unbiased statistics or independent, unbiased information? In what has been referred to as a ‘Data Deluge’, economies and societies are now swamped with all manner of fragmented statistics and indicators, many of them of poor quality. So at a time when modern economic and societal interactions are growing increasingly complex, there are many contradictory data available but relatively little good quality, integrated analyses.
    Consequently, there is a legitimate and useful role for the modern NSI in providing unbiased, useable information rather than pure statistics; distilling the available statistics to reveal the truth - to bring alive the stories hidden within the data. This can be done by providing more comprehensive and improved contextual macro-analysis to illustrate the relevance of overall official business statistics. Greater emphasis must also be given to improving data accessibility; physically, conceptually and visually. At the micro-level, NSIs could provide individual respondent feedback in the form of tailored, firm specific performance metrics or micro-analysis. This would not only provide enterprises with valuable information but also reduce perceived respondent burden, thus incentivising respondent compliance.
  • The Energy Production Profile of a Large Number of Residential Co-generators: a Statistical Evaluation

    Authors: Antonio Pievatolo (CNR-IMATI), Raffaele Argiento (CNR-IMATI), Stefano Galli (CNR-ITC), Matteo Mariotto (CNR-ITC)
    Primary area of focus / application: Modelling
    Keywords: micro-CHP, aggregate energy production, lognormal holding times, time-varying covariates
    Submitted at 30-May-2012 09:39 by Antonio Pievatolo
    Accepted
    11-Sep-2012 16:30 The Energy Production Profile of a Large Number of Residential Co-generators: a Statistical Evaluation
    Data obtained from an extensive experimentation, originally designed to evaluate the comfort and consumption performances of a micro-cogeneration system in a residential context, provide valuable information on the usage pattern of these systems. Using these data, an on/off stochastic process of non-exponential holding times depending on time varying covariates is estimated. The fitted model provides a building block for the estimation (with uncertainty bounds) of the daily profile of the energy produced by a large number of micro-cogenerators. To this end, a special Monte Carlo simulation technique, based on dependent samples, is devised to describe the energy production in a typical winter day.
  • Electric Load Forecasting in Italy by Exponential Smoothing with Covariates

    Authors: Kristina Lurz (University of Würzburg), Rainer Göb (University of Würzburg), Antonio Pievatolo (CMR-IMATI)
    Primary area of focus / application: Modelling
    Keywords: forecasting, exponential smoothing under covariates, trend, seasonality
    Submitted at 30-May-2012 09:43 by Kristina Lurz
    Accepted
    11-Sep-2012 16:50 Electric Load Forecasting in Italy by Exponential Smoothing with Covariates
    The lack of electricity storage solutions requires to provide reliable methods for forecasting the electric load. Electricity providers in Italy have a great interest in forecasting the electricity consumption of up to about 10 days in advance. The electricity consumption is influenced by various external factors, like the time of the day, the weekday and weather conditions. We make use of these factors in the prediction of the electric load by applying the method of exponential smoothing with covariates to hourly electric load data and compare the performance of different prediction intervals.
  • Some Marketing and e-Commerce Applications of the COM-Poisson

    Authors: Sharad Borle (Rice University)
    Primary area of focus / application: Modelling
    Keywords: COM-Poisson, count data, Marketing and e-Commerce
    Submitted at 2-Jun-2012 10:55 by Sharad Borle
    Accepted
    11-Sep-2012 12:10 Some Marketing and e-Commerce Applications of the COM-Poisson
    The two parameter generalization of the Poisson distribution has steadily been finding applications in diverse fields. This talk will introduce some of the applications that the COM-Poisson has been put to in the Marketing and ecommerce literature. The Poisson and Negative binomial distributions remain popular tools when it comes to applications involving count data in these fields. However, there have been applications that have benefitted from the property of COM-Poisson that allows for added flexibility to model under- as well as over-dispersed data.