Overview of all Abstracts

The following PDF contains the Abstractbook as it will be handed out at the conference. It is only here for browsing and maybe later reference. All abstracts as PDF

My abstracts


The following abstracts have been accepted for this event:


    Authors: Franco Pavese (National Institute for Research in Metrology, Torino, Italy)
    Primary area of focus / application:
    Submitted at 9-Sep-2007 17:07 by
    Replication of measurements and the combination of observations are standard and essential practices in metrology. A metrological –or testing– process of evaluating the uncertainty of the measurement results consists, in each laboratory, of basically three steps, using different methods to fulfil distinct purposes [1]:
    (a) When done on the same standard, to obtain the statistical features of the observations allowing to assess the repeatability of the value of the standard;
    (b) When done on the same standard, to obtain a measure of the effect on the total uncertainty of the variability of the influence parameters affecting the standard, including dependence on time, i.e. to assess the reproducibility of the value of the standard;
    (c) When done on several standards of the laboratory, to check if they have the same value or to establish the differences between their values, and to evaluate the associated uncertainty; i.e., to evaluate the accuracy of the values of the laboratory standards. This exercise can be called intra-laboratory comparison.

    When the exercise is performed for purpose (c) by comparing one (or more) standards provided by different laboratories, it is called inter-laboratory comparison. Past experience suggests that one should assume, as an a priori knowledge, that the comparisons are performed to detect bias.

    Bias is originating from the influence quantities, whose variability can show a non-zero mean and is also the source of the generally higher uncertainty obtained in ‘reproducibility conditions’ with respect to ‘repeatability conditions’.

    The paper is shortly recalling first the basic terms used in several written standards and international documents that not always are fully consistent each other and also show some evolutions of the concepts in the past decade, with the consequent possible confusion arising from the fact that not everybody is talking of the same things when they are assumed to. Then, is comparing several data models and is discussing their merits in taking (or not) into account the systematic effects, which are the prevailing reason of systematic errors in most metrology and testing measurements.

    [1] Pavese F and Filipe E 2006 Some metrological considerations about replicated measurements on standards Metrologia 43 419–425
  • How engineers learn statistics from motor cycle tires !

    Authors: J.J.M. Rijpkema (Eindhoven University of Technology, The Netherlands)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 01:03 by
    Design Based Learning focuses on an integrated approach to problem solving and engineering design, where students are stimulated to apply concepts and insights gained from mono-disciplinary courses in a multidisciplinary way. Its aim is to expand students' engineering competencies, while working on authentic problems using up-to-date CAE-tools. Design Based Learning offers opportunities to enhance the students' awareness of situations where Engineering Statistics can be relevant, as for example the design of experiments or the modeling and analysis of data may be part of the solution strategy.

    In this presentation I will explain the ideas behind Design Based Learning and its implications for our teaching of Engineering Statistics. I will illustrate this with details from a design based project we run for first year’s students in Mechanical Engineering at the Eindhoven University of Technology in the Netherlands. In this project tire-characteristics that are crucial for safe road handling of a motor cycle have to be determined experimentally from a full scale experimental setup. I will present experiences from students and lessons learned. Finally, I will discuss ways to enhance engineers’ statistical competencies through the use of design based projects in industry.

    Specifics: Related to the field of statistics education for engineers. The preferred form of presentation is an oral presentation of about 20 minutes (including discussion).
  • Supermarkets Sales and Retail Area: A Mixture Regression Model for Segmentation

    Authors: Ana Oliveira Brochado (1), Castro, Alexandra Carla (2) and Francisco Vitorino Martins (2)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 01:05 by Ana Brochado
    Sales in supermarkets, located across a country, can be explained by a wide range of
    factors as average price, store area for selling, service quality, regional purchasing
    power, population and competition.

    We intend to construct segments of stores (hypermarkets and supermarkets), identifying
    characteristics based on different responsiveness of sales to the different explanatory

    We use mixture regression models with a constant elasticity model (power function),
    estimating simultaneously the regression model and the assignment of each store to
    each segment. The number of segments is accessed by statistical indicators, as
    information and classification based criteria.

    Due to the high importance attributed by managers to the retail area elasticity, this
    variable has a special role in this analysis, named on the profiling of group members.
    Data were collected for all 106 stores belonging to national retail group, which sells
    three own retail brands. Some managerial implications were drawn.

    Keywords: Retail, supermarkets, sales, retail area, mixture regression models


    (1) Instituto Superior Técnico, Department of Civil Engineering and Architecture, Av. Rovisco Pais, 1049-
    001 Lisboa, Portugal, E-mail: abrochado@civil.ist.utl.pt

    (2) Universidade do Porto, Faculdade de Economia, Rua Dr. Roberto Frias
    4200-464 Porto, Portugal,E-mail: vmartins@fep.up.pt

    Specifics: poster presentation
  • Greedy Learning Algorithms and their Applications to Decision Trees

    Authors: Irad Ben-Gal and Niv Shkolnik (Tel-Aviv University, Tel-Aviv, Israel)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 10:38 by
    In this paper we considered the problem of classification by decision trees. In order to
    better address the inherent tradeoff between optimality and complexity we implement
    a revised version of the LRTA*, a known artificial intelligence algorithm. We show
    that such implementation leads to appealing properties of the classification tree.
  • Kriging-Based Sequential Inspection Plans for Coordinate Measuring Machines

    Authors: P. Pedone, D. Romano and G. Vicario
    Primary area of focus / application:
    Submitted at 10-Sep-2007 10:46 by
    In the last two decades kriging models, originally developed for geological applications, have gained increasing popularity in Computer Experiments as a tool for producing accurate predictions of the output of a deterministic computer code. Going against the flow and back to the origins, the paper shows how kriging models may also be effective in a physical experimental setting. Exploiting their recognized prediction capability, we use them to build sequential experiments to be applied to an engineering problem: the construction of inspection plans for checking the compliance of industrial parts to dimensional and geometric specifications on Coordinate Measuring Machines. The inspection plan specifies which points are probed and in which order. Since the economy of the process forces the sample to be small, the engineering objective is to accurately estimate deviations from nominal dimensions and shape by probing a few points only. As best accuracy/cost trade-off is also the objective of sequential designs, the inspection plan will be treated as a sequential experiment to be designed on-line. In the paper we present a number of case-studies, related to the check of form tolerances (straightness, circularity), whose related form error depends heavily on the extreme values of shape deviations. So, in the construction of the sequential design we consider both informative (maximum prediction variance) and problem-specific (search for extreme deviations) criteria. Performance of kriging-based plans is compared with that of the simple non sequential ones massively used in industrial practice (uniform, random, stratified) and with deterministic sequential methods in the engineering literature.
  • A comparison between two methods for measuring uncertainty based on the ISO Guides

    Authors: L. Deldossi and D. Zappa (Cattolica University of Milan, Milan, Italy)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 10:48 by
    In the last years the evaluation of the quality of the measurements is more and
    more frequently assessed by applying a new dedicated procedures defined in the ISO
    Guide (1995) - named Guide to the Uncertainty of the Measure (GUM) - based on the
    law of propagation of errors and on the Welch-Satterthwaite approximation. It is an
    alternative to the classical one defined in the ISO Guide 5725 (1994) based on the
    concept of repeatability and reproducibility obtained trough a variance component

    By means of some examples the two approaches are illustrated and compared to
    evaluate their limitations and their powers. Our proposal is to give some
    suggestions to the users about the method to be preferred according to their

    The question it is becoming relevant because process variation is decreasing due to
    technological improvements, therefore the measurement procedure has to be adequate.
  • Safety Improvement Using DMAIC Algorithm

    Authors: Michal Tkác; and Marek Andrejkovic; (University of Economics Bratislava, Kosice, Slovak Republic)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 11:09 by
    With this paper we intend to specify algorithm of realization and evaluation of safety improvement projects. We have used Six Sigma DMAIC methodology to solve the given, safety improvement problem. Individual stages of project realization are described in detail with designed methods, used in selected phases of the DMAIC. At the end of the paper, we present a case study from a realized safety improvement project in a U.S. company operating in metallurgical industry. In project realization, we have used methods like: integrated FMEA, expert opinion, statistical hypothesis testing, and computer simulations. At the end of the paper, we proposed how to evaluate safety improvement projects, using two criterions: technical and economical.

    KEY WORDS: Safety Improvement, Six Sigma, Expert Opinion, Project evaluation.
  • Managing Six Sigma projects - actively

    Authors: Michal Tkáč and Štefan Lyócsa (University of Economics Bratislava, Kosice, Slovak Republic)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 11:10 by
    Using real options theory, we had modeled a decision process about quality improvement investments under uncertainty. We have created a generic model adapted to Six Sigma quality improvement methodology, which values the managerial flexibility of running improvement projects. By means of a case study, we have shown how to employ our model and have shown the differences compared to traditional valuation technique, the net present value. At the end of the paper, we have shortly discussed the possible implication of real options theory in regard to cost of quality models.

    KEY WORDS: Project management, Six Sigma quality improvement, Real options
  • Economical aspects of training Six Sigma

    Authors: Michal Tkác; (University of Economics Bratislava, Kosice, Slovak Republic)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 11:11 by
    Costs of training, length of the training program, benefits from training, number of trained employees and other factors embody economical aspects of training Six Sigma. Mix influence of these factors and their management is the main topic of our paper. We are offering an analysis of a proposed business model of training Six Sigma Green Belts, which argues well-known questions of economical aspects of training programs. We will discuss how to measure the performance of training activities: the efficiency of investments into training of employees (from the financial as well as from qualitative perspective). How, our business model propagates the training of Six Sigma Green Belts and finally, how can we solve some antagonistic problems, when designing training programs.

    Key words: Education, Six Sigma Training, Return On Investment.
  • Brand Loyalty and Mixture Regression Models: Segmenting Customers in Jeans Market

    Authors: Ana Oliveira Brochado (1), Francisco Vitorino Martins (2) and Paula Cristina Rodrigues (3)
    Primary area of focus / application:
    Submitted at 10-Sep-2007 11:23 by Ana Brochado
    Brand value (BV) has been widely studied in recent years. The popularity of this subject
    is mainly due to its importance concerning strategic decisions as: differentiation,
    profitability and competitiveness of organizations, namely those evolving in an
    industrial environment.

    We intend to analyse loyalty - a brand-value effect, and test if the intangible value of a
    brand is (empirically) an effective determinant of BV, or, instead, if the only
    determinants of BV are tangible variables, as quality or notoriety.

    As the fashion industry (jeans) and their customers are very sensible to intangible
    factors, we use as explanatory variables brand personality and store image, despite the
    most classical determinants were brand personality and perceived quality.

    To evaluate the relevance of the intangible factors in explaining brand loyalty we
    consider the latent variable personality, witch were constructed based on the 15 items,
    proposed by the seminal article of Aaker.

    We intend to construct segments of customers based on brand loyalty and that present
    different responsiveness to the explanatory variables. As the dependent variable - brand
    loyalty - is a binary variable, we use a logistic mixture regression model.

    We collected information from 500 customers when they were shopping. We study 5 of
    the main brands of the jeans industry. Scales used in the questionnaire were the first
    factor analysed before the mixture study. Some managerial implications were drawn.

    Keywords: Brand-Value, Brand Loyalty, Brand Personality, Brand Quality, Fashion Industry, Jeans, Logistic Mixture Regression

    (1) Instituto Superior Técnico, Department of Civil Engineering and Architecture, Av. Rovisco Pais, 1049-
    001 Lisboa, Portugal, E-mail: abrochado@civil.ist.utl.pt

    (2) Universidade do Porto, Faculdade de Economia, Rua Dr. Roberto Frias
    4200-464 Porto, Portugal,E-mail: vmartins@fep.up.pt

    (3) Universidade Lusíada do Porto, Faculdade de Ciências Económicas e de Empresa, Rua Dr. Lopo de Carvalho,4369-006 Porto, Portugal, E-mail: 23010380@por.ulusiada.pt

    Specifics: poster presentation