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:

  • Identifying the Change Time of the Rate Parameter of a Multistage Poisson Count Process

    Authors: Mehdi Davoodi (Sharif University of Technology), Seyed-Taghi Akhavan-Niaki (Sharif University of Technology), Elnaz Asghari Torkamani (Amirkabir University of Technology)
    Primary area of focus / application: Process
    Keywords: Multistage Processes, INAR, Poisson Count Process, Change Point Estimation, MLE
    Submitted at 24-May-2013 09:51 by Mehdi Davoodi
    17-Sep-2013 16:55 Identifying the Change Time of the Rate Parameter of a Multistage Poisson Count Process
    Most often, researchers face count processes in today’s manufacturing environments. Moreover, multistage processes have attribute data that must be controlled. Estimating a process change time would simplify the search efforts to identify special cause(s). To provide a contribution in dealing with multistage count processes, this paper presents a model based on the first-order integer-valued auto-regressive INAR(1). Due to diversity of problem types, the model is developed considering Poisson marginal distribution. In the proposed model, Newton’s method is used to approximate the rate parameter, and the maximum likelihood method (MLE) is used to estimate the out-of-control sample and the out-of-control stage.
  • The Practical Use of Experimental Design in the Development of New Lubricants

    Authors: Maggie Wenham (Shell Global Solutions (UK))
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Experimental design, product development, product testing, choice of experimental units
    Submitted at 24-May-2013 11:12 by Maggie Wenham
    17-Sep-2013 10:10 The Practical Use of Experimental Design in the Development of New Lubricants
    The development of new lubricants involves a great variety of testing; from screening tests on new formulations in the lab to testing in vehicles or machinery under real-life operating conditions. Statistical design of experiments is used at all stages of the development process. This talk outlines the main types of design used at each of four stages: laboratory screening, lab-based emulation of service conditions, testing in vehicles/machinery and lastly, field trials.
    The general considerations of experimental design are discussed in relation to experiments at each stage, especially the challenges of keeping nuisance factors constant as you move closer to the real-life situation. Examples of Latin-square and balanced incomplete block designs are discussed in the context of laboratory tests emulating service conditions, using time factors as the blocking terms. Examples of designs for engine and vehicle tests are also given where particular attention is drawn to the issues of ensuring that the experimental units are properly identified so that the correct error term is used in the analysis. Finally, the ultimate challenge of conducting fleet trials is discussed, where the conditions are far from the idealised conditions in a laboratory environment.
  • Economic Aspects of Chain Sampling Plan under Inspection Errors

    Authors: George C. Mytalas (Athens University of Economics and Business. Dept of Statistics), Stelios Psarakis (Athens University of Economics and Business. Dept of Statistics)
    Primary area of focus / application: Quality
    Keywords: Chain sampling plans, Total sampling cost, Renewal process, Inspection error
    Submitted at 24-May-2013 13:13 by Stelios Psarakis
    In this paper, we propose a new chain sampling plan as an extension of ChSP-1 chain sampling plan introduced by Dodge (1977). We develop a mathematical model of this plan and obtain performance measures such as operating characteristic function, average total inspection and average outgoing quality considering constant inspection errors. We derive the total cost for the specific sampling plan that includes inspection costs, reworks, and defective items returned by the customers from accepted lots. First, the concept of a renewal reward process is employed to obtain the long-term net profit of the inspection policy. The second and the more important objective of this paper is to derive the minimum cost function for the chain sampling plan.
  • Parameter Estimation in Differential Equation Networks

    Authors: Dirk Surmann (TU Dortmund, Fakultät Statistik), Sebastian Krey (TU Dortmund, Fakultät Statistik), Uwe Ligges (TU Dortmund, Fakultät Statistik), Claus Weihs (TU Dortmund, Fakultät Statistik)
    Primary area of focus / application:
    Keywords: modelling, energy network, Low Frequency Oscillation, mechanical harmonic oscillators, differential equations, parameter estimation
    Submitted at 24-May-2013 18:15 by Dirk Surmann
    16-Sep-2013 14:45 Parameter Estimation in Differential Equation Networks
    The European electrical transmission system is operated increasingly close to its operational limits due to market integration, energy trading and the increased feed-in by renewable energies. For this reason it is necessary to analyse that part of energy that permanently oscillates through the electrical transmission system with a low frequency. These so called Low Frequency Oscillations are described and analysed within a smaller electrical system, the New England Test System, which guarantees a convenient handling. The analysis results in a new model which describes each node of the transmission system over mechanical harmonic oscillators. As in a true transmission system, the harmonic oscillators are connected over mechanical components according to the transmission lines of the electrical system. A simpler model allows faster predictions for the behaviour of the Low Frequency Oscillation if the topology of the network changes. Furthermore, a simple model can simulate much bigger networks at a fraction of time.

    All parameters of the model which bases on a system of differential equations are estimated from noisy data. The data is generated by a much more complex simulation system used at the Institute of Energy Systems, Energy Efficiency and Energy Economics of TU Dortmund University. The data is called noisy, because the simpler model will not describe all information from the energy simulation software. The talk works out the quality of the mechanical harmonic oscillator approach with respect to the complex energy simulation software.
  • Two-sample t-Test under Serial Dependence

    Authors: A.Ezgi Yılmaz (Hacettepe University), Serpil Aktaş (Hacettepe University)
    Primary area of focus / application: Other: Hypothesis testing
    Keywords: two-sample tests, t-test, serial dependence, autocorrelation
    Submitted at 25-May-2013 17:33 by Serpil Aktas
    17-Sep-2013 09:40 Two-sample t-Test under Serial Dependence
    The classical two-sample t-test assumes that observations are independent. Violation of this assumption might lead to inaccurate results and incorrectly analyzing these data leads to erroneous statistical inferences. However, in real life applications, such as in environmental, biological or ecological studies, data are often recorded over time and serial correlation is unavoidable. In this study, a new autocorrelation corrected method is proposed for the aforementioned case.
  • Generalized Maximum Entropy Approach to Unreplicated Factorial Experiments

    Authors: Mustafa Murat Arat (Hacettepe University), Serpil Aktaş (Hacettepe University)
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Unreplicated factorial designs, Entropy, GME, Factorial designs
    Submitted at 25-May-2013 17:33 by Serpil Aktas
    17-Sep-2013 10:50 Generalized Maximum Entropy Approach to Unreplicated Factorial Experiments
    In the initial stage of developing an industrial process, experimental studies based on factorial designs are often used to determine which factors among a number of factors have an effect on the response variable. A large number of factors somehow may arise and a number of runs that grows exponentially with the number of factors to be analyzed. Therefore, researchers often design unreplicated factorial experiments. Furthermore, considering the some limitations such as cost, time and data limitations, unreplicated factorial desings can be adopted to reduce the number of runs. But, using OLS method to analyze unreplicated experimental data results in zero degrees of freedom for error term in regression analysis. Generalized maximum entropy (GME) which is a method of selecting among probability distributions to choose the distribution that maximizes uncertainty or uniformity remaining in the distribution, subject to information already known about the distribution, is an alternative way of analyzing the unreplicated experiments. In this paper, GME is applied to a real data set and results are compared to the alternatives with respect to their efficiencies.