ENBIS-11 in Coimbra

4 – 8 September 2011 Abstract submission: 1 January – 25 June 2011

My abstracts


The following abstracts have been accepted for this event:

  • Applied Statistics as Competitive Advantage

    Authors: Igor Barahona, Alex Riba
    Affiliation: Department of Statistics and Operations Research at the Technical University of Catalonia
    Primary area of focus / application: Business
    Keywords: Applied statistics , Scale to rank companies , Key drivers , Competitive advantage
    Submitted at 6-May-2011 11:39 by IGOR BARAHONA
    6-Sep-2011 13:10 Applied Statistics as Competitive Advantage
    The technological evolution over the last 50 years has brought to the business environment more powerful computers for making more complicated statistical calculations. Nowadays, almost all companies use some kind of statistical technique based on either spreadsheets or specialized software to make better decisions.

    Although having better technological tools, Applied Statistics in business generally has a local impact; because it is only used to make decisions at departmental level. Furthermore, neither it has an impact at strategic level, nor it is an important part of company’s competitive advantage.

    How is Applied Statistics expanded within the company? The first step is to find out its actual degree of use. Considering this, the present lecture suggests to design a scale to rank the companies according to the use of statistical methods. The lowest value at the scale is for companies that don´t use any statistical technique and the highest is for those that make statistical practice a distinctive competitive advantage. At the middle are developing companies, who are improving on daily basis.

    Additionally, 5 key drivers to raise statistic’s use are introduced: Identification of competitive advantages, senior management support, systematic thinking, the use of data, and inventories of statistical techniques. Any company can ascend in the scale by improving its performance in each of these 5 key drivers.

    Finally, results obtained with a sample of Catalan Companies are presented and discussed, and the weight of each driver is calculated. Based on the two with the highest weight, an action plan is proposed.
  • Using Simulation for Measurement Uncertainty and other Industrial Applications

    Authors: Birger Stjernholm Madsen
    Affiliation: Novozymes
    Primary area of focus / application: Metrology & measurement systems analysis
    Keywords: Simulation , Normal Distribution , Skewness , Kurtosis , Measurement Uncertainty , GUM
    Submitted at 6-May-2011 14:25 by Birger Madsen
    5-Sep-2011 16:35 Using Simulation for Measurement Uncertainty and other Industrial Applications
    Simulation is in my opinion an overlooked tool within the field of industrial statistics. This may be due to the fact, that this was previously a cumbersome task, usually done using programming in a traditional (3rd generation) programming language or maybe a software tool like SAS or MATLAB (still excellent tools for simulation!). However, simulation nowadays can easily be done using more user-friendly statistical packages like JMP or Minitab, or even Excel in simple cases.

    In this presentation, I will illustrate some examples of simulation, e.g.:

    1. Illustrating the shape of the normal distribution for different sample sizes.

    The normal distribution assumption is crucial in many contexts, e.g. when calculating process capability indices.
    Many people are surprised, when an “obvious non-normal distribution” is tested for normality and accepted, when the sample size is small... On the other hand, many people are surprised, when an “obvious normal distribution” is tested for normality and rejected, when the sample size is large...

    2. Calculating limits for skewness and kurtosis, when testing for normality.

    A useful tool is calculation of the skewness and kurtosis. These descriptive statistics also provide information on how to “fix” a non-normal distribution, e.g. by a log-transformation (large skewness) or removing outliers (large kurtosis).
    The big question is: How large deviations from 0 are acceptable? Simulation can be used to indicate, when the asymptotic results are valid and to calculate maximum acceptable deviation from 0 for both skewness and kurtosis for small sample sizes.

    3. Using simulation in calculation of measurement uncertainty.

    Measurement uncertainty is often calculated in an uncertainty budget using the “Bottom up” (or GUM) approach. The assumptions of linearity and normal distributed output in are crucial. If these assumptions are dubious, the reported uncertainty may be of little value.
    I suggest an alternative approach, in which simulation replaces calculation of the combined and expanded standard uncertainty. This gives an overview of the actual distribution and a supplement to the traditional approach!
  • Estimation of a Normal Process Variance from Measurements with Large Round-Off Errors

    Authors: Benson-Karhi Diamanta and Schechtman Edna.
    Affiliation: the Open University of Israel and Ben-Gurion University of the Negev, Israel
    Primary area of focus / application: Metrology & measurement systems analysis
    Keywords: calibration curve , discrete scale , MLE , Sheppard correction , SPC , systematic error
    Submitted at 8-May-2011 14:04 by Diamanta Benson-Karhi
    5-Sep-2011 16:55 Estimation of a Normal Process Variance from Measurements with Large Round-Off Errors
    Measurements are sometimes affected by excessively large round-off errors. Small
    rounding-off may safely be ignored for purposes of statistical inference. The
    importance of round-off is determined by the ratio between the standard deviation σ
    and the instrument's scale step h, given by δ =σ / h . In this study we estimate σ
    when δ is small (δ < 0.5 ) using a calibration model. The calibration estimators are
    compared with a naïve estimator, Sheppard's correction and the maximum likelihood
    estimator (MLE), using simulation. We find that the calibration estimator can
    significantly improve the estimation in terms of MSE and bias, especially under
    circumstances where the other methods are not accurate or cannot provide a
  • Acceptance Sampling Plans by Non-Gaussian variables with Simultaneous Specification Limits

    Authors: Elisabete Carolino, Isabel Barão
    Affiliation: ESTeSL, IPL, Portugal / DEIO, FCUL, Portugal
    Primary area of focus / application: Process
    Keywords: acceptance sampling , non-Gaussian random variables , asymmetry , simultaneous specification limit , quality control
    Submitted at 9-May-2011 12:24 by Elisabete Carolino
    6-Sep-2011 11:05 Acceptance Sampling Plans by Non-Gaussian variables with Simultaneous Specification Limits
    In the quality control of a production process (of goods and services), from a statistical point of view, focus is either on the process itself with application of Statistical Process Control, or on its frontiers, with application of Acceptance Sampling (AS) – studied here – and Experimental Design.
    AS is used to inspect either the output process – final product – or the input – initial product. The purpose of AS is to determine a course of action, not to estimate lot quality. AS prescribes a procedure that, if applied to a series of lots, will give a specified risk of accepting lots of given quality. In other words, AS yields quality assurance. An AS plan merely accepts and rejects lots, considering sampling information.
    The AS by variables is based on the hypothesis that the observed quality characteristics follow a known distribution, namely the Gaussian distribution (classical case of the AS by variables – treated in classical standards). This is sometimes, however, an abusive assumption, that leads to wrong decisions.
    AS for non-Gaussian, mainly asymmetrical variables, is thus relevant. When we have a non-Gaussian distribution we can build specific AS plans associated with that distribution. In acceptance sampling we can built plans with a single specification limit (upper or lower) or simultaneous specification limits - studied in this work. In the literature there are few studies on simultaneous limits on acceptance sampling.
    In this work we will address the problem of determining acceptance sampling plans by variables with simultaneous specification limits for Gamma distributions (Exponential, in particular), the results being compared to the Gaussian case.

    ANSI/ASQC Z1.9-2008 2008. Sampling procedures and tables for inspection by variables for percent nonconforming. ASQ, Milwaukee, WI, EUA.
    ANSI/ASQC Z1.9-1980 1980. Sampling procedures and tables for inspection by variables for percent nonconforming. ASQ, Milwaukee, WI, EUA.
    Carolino, E., Casquilho, M., Barão, M. 2007. Amostragem de aceitação para uma variável assimétrica: a Exponencial. Actas do XIV Congresso Anual da Sociedade Portuguesa de Estatística, 281-292.
    Duncan, A. J. 1986. Quality Control and Industrial Statistics. 5.th edition. IRWIN, USA.
    Figueiredo, F. O. 2002. Controlo Estatístico da Qualidade e Métodos Robustos. Dissertação para obtenção do grau de Doutor em Estatística e Investigação Operacional, Especialidade em Probabilidades e Estatística. FCUL, Portugal.
    Figueiredo e Gomes. 2002. Transformação de Dados em Controlo Estatístico de Qualidade. Novos Rumos em Estatística. Sociedade Portuguesa de Estatística, 235-245.
    Box, G.E.P. and Cox, D.R. 1964. An Analysis of Transformations. Journal of the Royal Statistical Society, 211-252.
    Gomes, M. I., Figueiredo, F., Barão, M. I., 2010. Controlo Estatístico da Qualidade, 2ª edição revista e aumentada, Edições SPE.
    Grant, E. L. and Leavenworth, R. S. 1996. Statistical Quality Control, 7.th edition. McGraw-Hill, New York, NY, EUA.
    Guenther, William C. 1977. Sampling inspection in statistical quality control. First published. Whitstable Litho Ltd, GB.
    ISO 3951-1:2008. 2008, ISO standard.
    Levinson, W. 1997. Watch out for non-normal distributions of impurities. Chemical Engineering Progress. May, 70-76.
    Montgomery, D. C. 2004. Introduction to Statistical Quality Control, 5.th edition. John Wiley and Sons, New York, NY, EUA.
    Schilling, E. G. and Neubauer. D. V. 2009. Acceptance Sampling in Quality Control. 2nd edition. Chapman & Hall/CRC, USA.
    Wetherill, G. B. and Brown, D. W. 1991. Statistical Process Control. Chapman and Hall, London, UK.
  • Customer satisfaction and quality of service: analysis of coherence and latent variables for the University ’s canteen service

    Authors: Rossella Berni- Department of Statistics “G.Parenti”, University of Florence, Italy; Alessandra Durio- Department of Statistics and Applied Mathematics “D. de Castro”, University of Torino, Italy
    Primary area of focus / application: Business
    Keywords: Customer satisfaction; , Confirmatory factorial analysis , Structural equation models , Log-Linear models
    Submitted at 9-May-2011 13:23 by Rossella Berni
    Accepted (view paper)
    5-Sep-2011 17:01 Customer satisfaction and quality of service: analysis of coherence and latent variables for the University ’s canteen service
    In this paper we shall focus our attention to the concepts of customer satisfaction (CS) and quality of service (QS) in order to verify the validity of the measures involved in our case study. Undoubtedly, in the literature a long debate has been developed to define and to distinguish between these two measures. Starting from the SERVPERF and SERQUAL models, confirmatory analysis and structural equations models are applied in order to analyze the satisfaction of young students daily using the University’s canteens of Florence.
    In order to identify the latent variables and the measurement groups, we resort to a preliminary confirmatory factorial analysis that gives us also some information on the covariance between the latent variables themselves. The models of structural equations are identified; the systems of structural equations are defined using the standard notation stated by the Lisrel methodology.
    According to the evaluation expressed by the students interviewed on some crucial aspects of the perceived quality, we ratify the causal connections between the latent variables of the model and, at the same time, we provide for a dimension of quality perceived by the customers.
    Furthermore, the non-responses and the coherence of respondents is analysed through hybrid log-linear models in order to evaluate the burden of missing values and the individual response in two occasions of the same interview: ex-ante and ex-post.
  • Using Automated Adaptive Experimentation to Achieve Constrained Optimisation

    Authors: Chris Marley, Dave Woods, Sue Lewis
    Affiliation: University of Southampton, UK
    Primary area of focus / application: Design and analysis of experiments
    Keywords: Designed experiments , Expected improvement , Gaussian process , Latin hypercube
    Submitted at 9-May-2011 17:47 by Chris Marley
    6-Sep-2011 11:45 Using Automated Adaptive Experimentation to Achieve Constrained Optimisation
    Modern chemical processes are often highly multivariate, in terms of both input and output variables, and there may be complex relationships between the responses and predictors. There is therefore a need to identify robust and optimum operating conditions in terms of both desirable outputs and constraints on, for example, by-products.

    Recent advances in technology mean that it is sometimes possible to automate a series of chemical reactions to be performed one at a time, without the need for any reprogramming of machinery or further intervention of a chemist. This is known as “continuous flow mode”, and enables these systems to be left running continuously (for instance overnight), hence reducing costs.

    To enable such systems to be fully utilised for process optimisation, the reactions to be performed need to chosen adaptively and automatically. We propose a flexible approach for choosing these reactions based on the Expected Improvement criterion, often used for computer experiments. This general framework allows us to adaptively select design points with the goal of optimising an objective function subject to constraints. We present a simple example incorporating the choice of initial design, the adaptive selection of design points, model fitting and sensitivity analysis.