ENBIS-13 in Ankara

15 – 19 September 2013 Abstract submission: 5 February – 5 June 2013

My abstracts

 

The following abstracts have been accepted for this event:

  • Active Learning Exercise for Teaching the Planning and Documentation of a Series of Experiments

    Authors: Jacqueline Asscher (Kinneret College on the Sea of Galilee)
    Primary area of focus / application: Education & Thinking
    Keywords: Teaching statistics, Active learning, DOE, Documentation
    Submitted at 15-Apr-2013 16:32 by Jacqueline Asscher
    Accepted (view paper)
    18-Sep-2013 09:20 Active Learning Exercise for Teaching the Planning and Documentation of a Series of Experiments
    Some aspects of applying statistical design of experiments (DOE) are very familiar. For example, for each experiment we need to choose a design, choose factors, choose levels of factors, both those controlled in the experiment and those held constant, choose a response or responses, as well as to take into account the logistics inherent in the process and the experiment and the variation in materials and conditions.
    This active learning exercise focuses on the less tangible aspects of applying DOE, including how to document the process of designing a series of experiments, how to combine DOE with other approaches, such as testing a "best case" version to prove feasibility, and how to identify assumptions related to the process and to determine how these assumptions dictate our experimental strategy. In particular, in the case study in the exercise we address the common situation where a cheap or fast surrogate process is used in experimentation, and we consider two different scenarios for the assumed relationship between the surrogate process and the real process.
    The exercise was used as a stand-alone session at a conference with a mixed group of about 200 participants with varying levels of knowledge of DOE. It proved a useful starting point for discussion by participants, highlighting how differences in choice of experimental strategy reflect individual preferences and the influence of organizational culture. In addition, it was a non-intimidating advertisement for statistical DOE. In this paper I discuss both how to plan and document a series of experiments and how to construct and conduct such an exercise. The exercise itself is included.
  • Environmental Application of a Compositional CUSUM Control Chart

    Authors: Marina Vives-Mestres (Universitat de Girona), Josep-Antoni Martín-Fernández (Universitat de Girona), Josep Daunis-i-Estadella (Universitat de Girona)
    Primary area of focus / application: Other: Talk withdrawn
    Keywords: Compositional data, Log-ratio, CUSUM, Environmental monitoring
    Submitted at 15-Apr-2013 16:39 by Marina Vives-Mestres
    Accepted
    Control charts, which were initially designed for monitoring industrial processes, are now applied to other areas such as public-health or health-care surveillance and environmental monitoring. We focus our study on the application of control charts to environmental monitoring (e.g. air or water quality) as measurements are mostly concentrations, i.e., compositional data.
    Compositions are vectors of positive elements describing quantitatively the parts of some whole, which carry exclusively relative information between the parts. Their sample space is the simplex and specific statistical methods are necessary because its particular geometry. This approach is based on the log-ratio methodology.
    CUSUM control charts enable monitoring deviations from the process mean and have proven to be very effective in detecting small process shifts. In recent years, the CUSUM scheme has been applied to analyse changes in air pollution levels associated with the introduction of traffic management schemes. CUSUM reveals to be a simple tool that helps in decision making.
    We analyse the effect of the introduction of a bus lane on a congested route in Central London using as indicator the hourly measurements of carbon monoxide (CO) concentrations (mg m-3). The reference mean and standard deviation are calculated from the pre-implementation period and used to determine whether the traffic management scheme has an effect on the indicator of traffic pollution (CO).
    The log-ratio approach is used to treat the data as parts of a whole, thus analysing the relative proportion of the CO concentration. We compare our results with the classical approach and enhance the advantages and drawbacks of the proposed method in the area of application.
  • From Safety Margins to Stochastic Models

    Authors: Michael Ashcroft (Inatas AB/Uppsala University)
    Primary area of focus / application: Reliability
    Keywords: Control, Bayesian networks, Sensor failure, Risk analysis, Decision theory
    Submitted at 15-Apr-2013 16:59 by Michael Ashcroft
    Accepted
    Many industries are accustomed to working with a ‘Safety Margin’ based sensor failure analysis. In this paper, I will use a simple example from Waste Water Treatment to discuss the value that results from moving from such an approach to the use of Bayesian network based models (BNs).

    I will examine how BNs can capture any information present in pre-existing safety margin methods. Then I shall discuss the merits that arise from working with a stochastic model rather than a binary ‘in-control’ variable. Special attention will be placed on the opportunity for more sophisticated risk analysis and the extension to decision theoretic models.
  • Causality Relationship Between R&D Expenditure and Economic Growth

    Authors: Perin Ünal (Middle East Technical University)
    Primary area of focus / application: Mining
    Keywords: Granger causality test, Research and development, R&D expenditure, Economic growth, Industrial economics
    Submitted at 15-Apr-2013 17:24 by Perin Ünal
    Accepted (view paper)
    17-Sep-2013 17:50 Causality Relationship Between R&D Expenditure and Economic Growth
    Analysing causal relationships has been a major area of interest in a wide range of application areas. Granger causality has been widely used to identify the causal relationships in multivariate regression models for time series. In this paper Granger Causality test is utilized to identify the causality relationship between research and development expenditures and economic growth in manufacturing industry in Turkey. This paper delivers findings in Turkish manufacturing industry and guides researchers to future thinking and research in developing countries.
  • Estimating Flight Delays using Airport Network Information

    Authors: Nilgün Ferhatosmanoğlu (THK University), Mehmet Güvercin (Bilkent University)
    Primary area of focus / application: Mining
    Keywords: Flight delay forecasting, Time series, Regression, Clustering analysis, Hub score, Graphs and networks
    Submitted at 15-Apr-2013 18:12 by Nilgün Ferhatosmanoğlu
    Accepted
    16-Sep-2013 12:35 Estimating Flight Delays using Airport Network Information
    Flight delays are a major cause of delay propagation in airport network and are a major source of costs for airlines and passengers. Accurate prediction of flight delays is essential for both optimization of airline operations and airport capacity planning that directly affects customer satisfaction. It is a challenging problem as the underlying factors are usually not a priori known. In this study, we introduce a methodology for forecasting flight delays that integrates airport network information with time series forecasting models. We first introduce a graph structure that represents the interaction network of airports. We define graph-theoric measures for each airport on this interaction graph by making an analogy with the methods in social network literature. We cluster the airports using the extracted features, and use the clusters for a joint modeling for flight delay estimation. The proposed estimation approach incorporates the topological features of the airports into their time-series models. In particular, we integrate the REG-ARIMA forecasting model with a clustering of the airports using their graph topological features. We perform experiments on a data set consisting of 6.5 million commercial flights. The results confirm that the proposed method produces more accurate forecasts than the existing methods in the literature.
  • Modelling of Donor Careers for Managerial Decision Making

    Authors: Mart Janssen (UMC Utrecht, str 6.131), Onno Verhagen (Sanquin Blood Supply Foundation), Ellen van der Schoot (Sanquin Blood Supply Foundation)
    Primary area of focus / application: Business
    Keywords: decision making, simulation, survival analysis, prediction
    Submitted at 16-Apr-2013 09:49 by Mart Janssen
    Accepted
    17-Sep-2013 10:10 Modelling of Donor Careers for Managerial Decision Making
    Background
    In The Netherlands pregnant women have their blood examined around week 12 of pregnancy for RhD typing, and the RhD-negative women are tested again at week 27 to type the foetus. Each year 27,000 women are found RhD negative, from which 60% are carrying an RhD-positive child. To these women anti-D injections are administered at week 30 and within 48 hours to prevent the pregnant mother from producing antibodies against the RhD-antigen. These antibodies can cause rhesus disease, when the mother becomes pregnant again from an RhD-positive child. Before the introduction of the anti-D immunoprophylaxis, this disease was the major cause of perinatal death. In the Netherlands the anti-D prophylaxis is produced from the plasma from immunized volunteers. As this procedure is quite demanding for the donor and the number of naturally immunized women is shrinking, over the past years a strong decline in anti-D donors is observed.
    Aim
    To determine whether ageing of the ant-D donor population is the cause of the reduction of the size of the anti-D donor population, and what actions are required to stabilise the size of this population.
    Methods
    We estimated the length of the donor career using standard survival analysis techniques. Censoring (left and right) of donors had to be taken into account. Based on the observed survival a very simple but adequate model for donor retention was built. With this model we were able to predict the size and age distribution of the donor population over a large number of years.
    Results
    For repeat donors the length of the donor career is independent of the age of the donor. However, it can be shown that this length has changed considerably over time and that the current drop-out rate is considerably higher than it was in the past. This effect rather than ageing of the donor population is the causing the reduction in the number of donors. The developed model allows prediction of the size of the future donor population as a function of the annual donor drop-out rate and number of newly recruited donors.
    Conclusions
    The analyses show that with a relatively simple model it is possible to predict the size and age distribution of the Dutch anti-D donor population. This allows informing management on the efforts required in terms of donor recruitment and donor retention to maintain a pre-specified level of antibody production.