ENBIS-16 in Sheffield

11 – 15 September 2016; Sheffield Abstract submission: 20 March – 4 July 2016

My abstracts


The following abstracts have been accepted for this event:

  • RosettaHUB, Towards a Universal Platform for Open Statistical Computing in the Cloud

    Authors: Karim Chine (RosettaHUB)
    Primary area of focus / application: Mining
    Secondary area of focus / application: Education & Thinking
    Keywords: R, Cloud computing, Education, Collaboration, Reproducible statistical computing, Dashboards, SaaS for statistics
    Submitted at 31-May-2016 23:59 by Karim Chine
    14-Sep-2016 09:00 RosettaHUB, Towards a Universal Platform for Open Statistical Computing in the Cloud
    Fragmentation in the data science space reduces the productivity of statisticians and compromises their ability to share, collaborate and make their results reproducible. RosettaHUB is an innovative platform for open data science that reduces those frictions and offers statisticians a streamlined experience in their day-to-day interaction with tools, infrastructures, data and peers.

    R, Python, Julia, SQL, Scala, Spark, ParaView, etc. can be used simultaneously and collaboratively within an interactive Google-docs like environment. The platform acts as a broker and a marketplace for statistical applications, it makes it possible to create and share resources on any infrastructure including Amazon EC2, GCE, Azure and private clouds. RosettaHUB is fully programmable. Its statistical and numerical cloud-based capabilities are accessible from Excel and from all the mainstream environments used by statisticians on their Personal Computers including the R and Python command lines.

    Web site: www.rosettahub.com
  • Statistical Methods in Emotional Product Design Following the Kansei Engineering Model

    Authors: Lluis Marco-Almagro (Universitat Politecnica de Catalunya, BarcelonaTech), Xavier Tort-Martorell (Universitat Politecnica de Catalunya, BarcelonaTech)
    Primary area of focus / application: Other: Statistical methods in emotional design
    Keywords: Emotional design, Kansei engineering, Multivariate methods, Quantification theory type I, Regression analysis
    Submitted at 3-Jun-2016 18:33 by Lluis Marco-Almagro
    12-Sep-2016 11:30 Special Session: Kansei Engineering (Hands-on Session)
    When customers are questioned on what they want, a list of needs normally referring to functionality is obtained. However, customers do not usually explain their emotional needs, probably because they are not aware of having them or are unable to tell which they are.

    Some markets are currently so crowded of similar products in terms of functionality that adding an “emotional touch” can make a difference. How do designers create “emotional products”? They usually rely on their intuition, creativity and experience. But they also use different qualitative and quantitative methods to collect information on how products are perceived and used. One of this methods is the so-called kansei engineering (KE).

    Kansei engineering is a method for incorporating emotions in the product development phase. The main purpose is discovering which technical parameters of a product elicit the chosen emotions. The method was first proposed by Prof. Mitsuo Nagamachi in the 70’s and 80’s, but gained attention in this XXI century, in part due to work by Prof. Simon Schütte at Linköpings Universitet. KE studies are based on self-reporting emotional reactions with questionnaires (usually ratings on Likert or semantic differential scales). A set of different prototypes is shown to participants in the study, and ratings are given on elicited emotions. Each emotion acts as a response in a design of experiments.

    There is a large range of statistical tools commonly used in KE studies, mainly multivariate techniques and regression models. Data in KE studies have a great amount of variability, and as building prototypes is costly, there is always the attempt to discover a lot of things (having a lot of factors) but only a few runs in the experiment (probably too few!). All these issues pose interesting statistical challenges; in fact, kansei engineering is a discipline “in need of statistics”.

    This seminar will take the form of a workshop, where you will be asked to discuss and work with your colleagues. We will first cover the basic ideas behind kansei engineering studies, and present the model used to conduct them. After several examples, a real (simple) KE study will be prepared by participants in small groups. This small example will be used to discuss some statistical tools useful in KE. For instance, multivariate techniques for summarizing information, and for automatically detecting “crazy” participants, will be covered. As “customers” of KE studies are often designers, a great importance is placed on visual presentation of results. Quantification theory type I (QT1), a special version of regression analysis commonly used in KE, which makes interpretation of results easier when all independent variables are categorical, will also be explained (the ideas behind QT1 are, in fact, useful far beyond KE studies).

    When the seminar finishes, you will have a good understanding of what kansei engineering is, and how statistics can contribute to this field. Moreover, you will have learnt some “tricks”, such as QT1, that make statistical output easier to interpret to a broad audience.
  • A Locally Asymptotically Optimal Test for ARCH Models

    Authors: Joseph Ngatchou-Wandji (EHESP Rennes & Université de Lorraine), Tewfik Lounis (Université de Lorraine)
    Primary area of focus / application: Finance
    Secondary area of focus / application: Economics
    Keywords: Optimal tests, ARCH models, Power of tests, Contiguity
    Submitted at 4-Jun-2016 09:41 by Joseph Ngatchou-Wandji
    Accepted (view paper)
    13-Sep-2016 12:10 A Locally Asymptotically Optimal Test for ARCH Models
    We construct a locally asymptotically optimal test for discriminating between ARCH models within a large class. The test statistic is based on a new type of estimator. Its asymptotic properties are studied under the null hypothesis and under a sequence of local alternatives. The latter study is done by establishing a contiguity result and using the resulting LAN property. Our theoretical results extend existing ones, which are generally established under the assumption that the parameter vector is known. A simulation experiment shows that our test behaves well under the hypotheses considered. The results are also applied to a set of real data.
  • The Use of the Algebraic Method in the Design of Experiments

    Authors: Henry Wynn (London School of Economics)
    Primary area of focus / application: Design and analysis of experiments
    Secondary area of focus / application: Design and analysis of experiments
    Keywords: Grobner bases, Aliasing, Factorial design, Quadrature, Intervention
    Submitted at 4-Jun-2016 11:48 by Henry Wynn
    12-Sep-2016 14:30 The Use of the Algebraic Method in the Design of Experiments
    The use of Grobner bases in experimental design allows extensive study of the alias structure of both newly generated design and existing families of designs. It also shows how there is a Nyquist-type frontier
    showing what the limits of identifiability are for all polynomial models.
    A new application is via a type of duality which explains the use
    of certain type of design in quadrature. There are many open problems,
    foremost of which is to link the algebraic methods to optimal design, and their use in complex areas such as intervention in causal models defined by directed acyclic graphs (DAGs).
  • A Bayesian Short-Term Strategy for Site-Specific Wind Parameter Estimation

    Authors: Pasquale Erto (University of Naples Federico II), Antonio Lepore (University of Naples Federico II), Biagio Palumbo (University of Naples Federico II)
    Primary area of focus / application: Modelling
    Keywords: Markov chain Monte Carlo, Bayesian estimators, Wind speed distribution, Weibull distribution
    Submitted at 6-Jun-2016 13:04 by Antonio Lepore
    13-Sep-2016 16:20 A Bayesian Short-Term Strategy for Site-Specific Wind Parameter Estimation
    The economic profitability of a candidate site is strongly influenced not only by the wind speed, which is needed to define the turbine type to be installed, but also by its direction, which represents a dominant parameter in the wind-farm layout design. Wind potential assessment from short (e.g., one-month) sample can be severely poor if performed when the wind is not blowing from the prevailing direction(s). This paper proposes a Bayesian approach in order to hasten directional data analysis collected from an Italian candidate site. In particular, the wind direction is grouped in 8 sectors and modelled with the multinomial distribution. The Dirichlet distribution is chosen as prior and is calibrated on the historical data available at a neighbouring survey station. In particular, the prior distribution elicitation is based on the Fisher’s angular-angular association between directional data with speed value greater than 4 m/s , which are collected simultaneously from the candidate site and the neighbouring survey station. In expert opinion, such threshold value represents the least speed that effectively activates turbines. Then, the Bayesian approach proposed in Erto et al (2010), which involves MCMC (Markov chain Monte Carlo) method, is opportunely adjusted in order to furnish the estimates of the wind speed distribution grouped by angular sector. In such a way, prior information on wind features can be better incorporated in a more familiar way into the adopted prior distributions for the Weibull model parameters. The attained results based on the Mean Square Error show that the rose-plot based on the Bayesian estimates carried out from a 1-month sample is comparable to the actual 1-year one. Such analysis is proposed to cope with actual problems faced by renewable energy companies as encouragingly shown by an application to real anemometric data from a Southern Italian site.
  • Case Studies on PLS-Based Procedure for Ship Fuel Consumption Monitoring and Fault Detection

    Authors: Andrea D'Ambra (Grimaldi Group), Antonio Lepore (University of Naples Federico II), Biagio Palumbo (University of Naples Federico II), Luigi Vitiello (University of Naples Federico II), Christian Capezza (University of Naples Federico II)
    Primary area of focus / application: Modelling
    Secondary area of focus / application: Process
    Keywords: Partial Least-Squares regression, Fuel consumption monitoring, Fault detection, Hotelling control chart, Squared prediction error control chart, Vessel energy efficiency
    Submitted at 6-Jun-2016 19:18 by Luigi Vitiello
    13-Sep-2016 11:40 Case Studies on PLS-Based Procedure for Ship Fuel Consumption Monitoring and Fault Detection
    The new Regulation (EU 2015/757) of the European Union Council, based on the International Maritime Organization guidelines, urges shipping companies to adopt systems for monitoring, reporting and verifying CO2 emissions through the analysis of ship fuel consumption. The novel procedure presented in Bocchetti et al. (2015a) points out that Partial Least-Squares (PLS) methods could be successfully utilized (at the end of each voyage) by exploiting the navigation data overload collected on board by modern data acquisition systems. In this work, the procedure has been elaborated and implemented on Grimaldi Group’s twin cruise ships operating in the Adriatic Sea. Different case studies are presented to validate the procedure ability of alerting management for a possible change in ship performances as well as detecting anomalies in navigation variables (i.e., fault detection) that need further technological investigation. Moreover, the procedure is newly applied for estimating the fuel consumption reduction consequent to efficiency improvement operations (e.g., hull form optimization, hull cleaning and propeller polishing, ultra-smooth coating, propulsion efficiency improvement, engine maintenance operation, power plant efficiency improvement). This feature is particularly profitable for shipping companies and operators to quantify ship efficiency gain and therefore, to claim for carbon credits.