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:

  • Extreme Quantile Estimation and Satellite Positioning Applications

    Authors: J.Bureau, F.Gamboa, J.M.Loubes, S.Tarance
    Affiliation: HELILEO SA, Saint-Paul-lès-dax, France, and IMT, Paul Sabatier University, Toulouse, France
    Primary area of focus / application: Metrology & measurement systems analysis
    Keywords: GNSS , GPS , integrity , extreme quantile estimation
    Submitted at 27-May-2011 10:10 by Jérémie Bureau
    Accepted
    5-Sep-2011 12:30 Extreme Quantile Estimation and Satellite Positioning Applications
    Among the GNSS (Global Navigation Satellite System) applications currently used or in development, some of them require high performances in terms of precise positioning and reliability for safety of life. These critical performances are evaluated using statistical tools, and the problem of measuring the position accuracy or the system reliability (integrity) can be modeled as a quantile estimation. Specific safety of life applications, such as an airborne precision approach, require very high levels of quantiles (10−7) and lead us toward the use of statistical tools from ex- treme value theory, still unused in the GNSS field.

    Basics of this theory and extreme quantile estimation will be first presented. We will focus on the excesses over a threshold framework and particularly on the associated threshold selection with an automatic procedure dedicated to that step. Finally, we will outline a methodology for integrity analysis and some experimental results gathered during a helicopter test campaign.
  • Joining the Dots, How Workflow Journals are Connecting Our Thinking..

    Authors: Kate Llewellyn
    Affiliation: GSK
    Primary area of focus / application: Modelling
    Keywords: experimental design , custom design , supersaturated design , democratisation of data , workflow journal
    Submitted at 27-May-2011 14:21 by Kate Llewellyn
    Accepted
    6-Sep-2011 11:45 Joining the Dots, How Workflow Journals are Connecting Our Thinking..
    Experimental design has long been a cornerstone of chemical development, and the inventory of available designs has expanded significantly with the advent of computer-generated custom designs. Using the example of the development of a novel chemotherapeutic agent, this paper shows our current approach to API process development. A workflow journal was used to systemise and structure the data analysis and the scientific knowledge relating to this project. The democratisation of data in this way is key to leveraging its full value, and to reducing the impact of a silo-based approach to R&D.
    A range of classical and custom designs are used across the various chemical stages, including an alias-optimal design and a Bayesian supersaturated design. The journal integrates the statistical modelling with development, analytical and production data, telling our story more effectively at many levels to many different audiences.
  • Form Error Conformance Assessment

    Authors: Alistair B Forbes
    Affiliation: National Physical Laboratory
    Primary area of focus / application: Metrology & measurement systems analysis
    Keywords: form error , conformance assessment , measurement uncertainty , Gaussian processes
    Submitted at 27-May-2011 14:45 by Alistair Forbes
    Accepted
    6-Sep-2011 12:30 Form Error Conformance Assessment
    The principle task in coordinate metrology is to assess whether a manufactured part will meet its functional requirement from the measurement of the coordinates of points on the surface of the part [1]. In general, the functional requirement is encoded in a set of geometric tolerances that specify how far the part can depart from its ideal geometry, both in terms of size and shape, and still be able to perform. The assessment of form error requires algorithms to fit the ideal geometry to the measured coordinates. The residual distances of the points to the fitted geometry provide information about the form error, i.e., the departure from ideal geometry, and a decision to accept or reject the part is made on their basis. However, two factors cloud the decision making, the first is that the measurements are not exact, the second is that only a discrete representation of the part is available while its functional performance will generally depend on the compete surface. In this paper we consider how Kriging models can be used to improve form assessment methodologies, accounting for both measurement uncertainty and the discrete sampling effect.

    [1] A B Forbes and H D Minh, Form assessment in coordinate metrology, in Approximation Algorithms for Complex Systems, E H Georgoulis, A Iske and J Levesley, eds., pp 69–90, Springer, Heidelberg, 2011.
  • Human Error - Unavoidable ......

    Authors: Johan Batsleer
    Primary area of focus / application: Quality
    Keywords: human error , brain , failures , managing failures
    Submitted at 27-May-2011 19:02 by Johan Batsleer
    Accepted (view paper)
    7-Sep-2011 12:00 Human Error - Unavoidable ......
    Human error - unavoidable .....

    We all make mistakes. We must accept that humans make mistakes and will continue to do so. It is useless to make rules to forbid making mistakes.
    On the other hand, it also implies that individual measures will have a very limited effect. Measures on a larger scale, within processes and within the organization , will be needed.
    To comprehend human error, it is required to take a closer look inside the functioning of the human brain.
    It can help us to understand the mechanism of ‘human error’ in order to take necessary measures against human failure.

    Johan Batsleer, senior consultant, Amelior
  • The Total Median Statistic Revisited

    Authors: Fernanda Otilia Figueiredo and Maria Ivette Gomes
    Affiliation: Faculdade de Economia da Universidade do Porto and CEAUL; Universidade de Lisboa, FCUL, DEIO and CEAUL;
    Primary area of focus / application: Process
    Keywords: SPC , Control Charts , Robust Statistics , Order Statistics
    Submitted at 27-May-2011 19:18 by Fernanda Otilia Figueiredo
    Accepted (view paper)
    6-Sep-2011 16:05 The Total Median Statistic Revisited
    Despite the advantages of the use of the normal distribution, in particular in the context of statistical quality control, it is well known that most of the data sets from diversified industrial processes come from a non-normal distribution, with moderate to strong asymmetry and tails heavier than the normal tail. But even in potential normal situations we can have small disturbances in the data to the normality assumption, for instance, a contaminated normal distribution.
    Following Figueiredo and Gomes (2004, 2009), and taking in consideration some simulation results about the duality robustness/efficiency of the total median statistic, we analyze the performance of the total median chart to monitor the mean value of an industrial process, after an a priori application of a transformation to the original observations in order to change them into exponential values.
    This research is partially supported by FCT/OE, POCI 2010 and PTDC/FEDER.

    References:
    [1] Figueiredo, F. and Gomes, M. I. (2009). Monitoring Industrial Processes With Robust Control Statistics. Revstat - Statistical journal, 7(2), pp. 151-170.
    [2] Figueiredo, F. and Gomes, M. I. (2004). The total median is Statistical Quality Control. Applied Stochastic Models in Business and Industry, 20, pp. 339-353.
  • Will Managers Ever Embrace Statistics?

    Authors: Roland Caulcutt
    Affiliation: Caulcutt Associates, UK
    Primary area of focus / application: Business
    Keywords: statistics , management by fact
    Submitted at 27-May-2011 19:51 by Roland Caulcutt
    Accepted
    7-Sep-2011 12:20 Will Managers Ever Embrace Statistics
    Many statisticians would agree that managers need some understanding of basic statistics. Perhaps we can also agree that many managers do not have this understanding and do not appear to appreciate the need to acquire it. So, management decisions that could and should be based on data, are often supported only by opinion, with the opinions of more senior managers given greater weight.
    I believe the main reason why managers are not attracted to the study of basic statistics lies in the higher priority they give to their other needs as they struggle to acquire the wide range of skills that their craft requires. Indeed, when we list the many competencies that managers need it is hard to see how we can persuade them to give higher priority to managing by fact.