ENBIS Spring Meeting 2019 in Barcelona

13 – 14 June 2019 Abstract submission: 15 January – 20 May 2019

My abstracts


The following abstracts have been accepted for this event:

  • Statistical Engineering - My View

    Authors: Jonathan Smyth-Renshaw (Jonathan Smyth-Renshaw & Associates Ltd)
    Primary area of focus / application: Consulting
    Keywords: Statistical Engineering, Sexy, Un-sexy, Not Data Science
    Submitted at 27-Feb-2019 20:02 by Jonathan Smyth-Renshaw
    13-Jun-2019 15:10 Statistical Engineering - My View
    Statistical Engineering is nothing new. I have used Statistical Methods within Engineering for 30 years. So it is not new. My presentation will provide examples of the practical application of Statistical Engineering and demonstrate a structure for the application of Statistical Engineering which I have used over the last 30 years.

    Applications include give-away in the food industry, automotive production, climbing equipment and house insultation.

    Statistical Engineering is not Six Sigma or Lean or Big Data or Data Science. Statistical Engineering is the silver bullet to many problems as it considers the fundamental law of engineering and couples this with the discipline of statistics. Statistical Engineering is very powerful but it lacks the buzz work of Six Sigma, Lean or Big Data. The challenge is how can the 'unsexy' become 'sexy to business leaders'?
  • What’s the Role of Software in Statistical Engineering?

    Authors: Volker Kraft (JMP), Ian Cox (JMP)
    Primary area of focus / application: Other:
    Keywords: Multidisciplinary, Learning support, Knowledge encapsulation, Knowledge sharing, Empirical model building, Software
    Submitted at 15-Apr-2019 11:00 by Ian Cox
    13-Jun-2019 12:20 What’s the Role of Software in Statistical Engineering?
    Statistical Engineering seeks to solve tough problems with an holistic, multidisciplinary approach. This is a challenging agenda, requiring the help of software that can support learning, and encapsulate and make accessible the knowledge that results. This presentation aims to articulate the capabilities that software needs to do this adequately, and in so doing, suggests why new patterns of use may be required. These might be provided by more disciplined and consistent use of existing applications, or by software that does not yet exist. To make the scope manageable, we will focus on empirical model building using data, an important part of Statistical Engineering.
  • European funds local expenditures simulation: modelling validation and global sensitivity analysis

    Authors: Samuele Lo Piano (Universitat Oberta de Catalunya), Arnald Puy (Université Libre de Bruxelles), Emanuele Borgonovo (Università commerciale Bocconi), Andrea Saltelli (Universitetet i Bergen)
    Primary area of focus / application: Modelling
    Secondary area of focus / application: Economics
    Keywords: global sensitivity analysis, modelling, model validation, machine learning, econometrics, European funds
    Submitted at 29-Apr-2019 17:28 by Samuele Lo Piano
    13-Jun-2019 17:45 European funds local expenditures simulation: modelling validation and global sensitivity analysis
    Understanding the spending pattern of European funds is an issue of primary importance as to evaluate their effectiveness in generating benefits at variable spatial and time scales. This especially applies to funds that are remitted at the regional level, such as the ESIF (European Structural and Investment Funds), which represents the main financial tool the European Commission resorts to foster its cohesion policy. However, producing meaningful inference on the local effects yield by these funds faces an inevitable issue of time lagging: The payments from the European Commission are remitted only after the incurred expenses have been invoiced. Precisely, the only official figures available report on the dates when the EU payment were remitted to the local authorities. These figures may be delayed from the actual year on which expenditures were invoiced by the final beneficiaries to the local authorities. These figures are typically not available. Therefore, producing econometric inference on the benefits these ESIF locally produce on the ground may be cumbersome due to this very time lag.

    In this contribution, we present a model we developed to simulate the local expenditures on the basis of the observed reimbursement pattern from the European Commission to the European regions. We introduced a new measurement based on the distance of yearly cumulative figures – an index of regional specificity. This parameter along others sampled in plausible uncertainty ranges are used in Monte Carlo simulations to produce expenditure distributions.

    The availability of the actual expenditures invoiced to a selection of member state authorities allowed to validate our model through multi-scale benchmarking at different levels of granularity. This comparison resulted in improving the model by tuning the modelling parameters as to reduce the mismatch between the simulated and the actual incurred expenditures. Finally, global sensitivity analysis allowed to apportion the uncertainty in the modelled expenditure against the uncertainty in the input parameters.

    This study shows the potential of the combined use of machine-learning type of inference and global sensitivity analysis. The attained advantage is two-fold as it allows to: i) produce a robust modelling activity; ii) flag up the areas mostly affecting the output stochastic uncertainty.
  • Remaining Useful Lifetime Prediction: A case study in combining statistics and machine learning

    Authors: Alessandro Di Bucchianico (Eindhoven University of Technology), Yingjun Deng (Eindhoven University of Technology), Mykola Pechenizkiy (Eindhoven University of Technology)
    Primary area of focus / application: Modelling
    Secondary area of focus / application: Process
    Keywords: machine learning, stochastic degradation models, remaining useful life, maintenance
    Submitted at 2-May-2019 17:26 by Alessandro Di Bucchianico
    Accepted (view paper)
    13-Jun-2019 16:55 Remaining Useful Lifetime Prediction: A case study in combining statistics and machine learning
    Calculation of Remaining Useful Life (RUL) is an important new trend in maintenance. It is an essential element to go from diagnostic analytics to predictive or rather prescriptive analytics, to use data science terminology. In Industry 4.0 settings, we have to determine RUL in data rich environments with an abundance of sensor values and degradation indicators but few failures. In such settings machine learning techniques may be useful to get predictions when complex relations between sensor values exist that make statistical modelling not feasible. However, machine learning techniques typically fail short of providing indications of uncertainty. To combine the best of two worlds, we propose an approach that combines machine learning and statistics. Recurrent neural networks are used to provides predictions from degradation values, and we use stochastic degradation models to add uncertainty to the prediction.

    The proposed approach to RUL prediction will be illustrated with an industrial use case within the EU Horizon 2020 project Prophesy.
  • Statistical engineering challenges from big data in industry

    Authors: Biagio Palumbo (Univ. of Naples Federico II), Antonio Lepore (Univ. of Naples Federico II)
    Primary area of focus / application: Process
    Secondary area of focus / application: Modelling
    Keywords: shipping industry, manufacturing industry, high-speed rail industry, statistical process control, functional data analysis
    Submitted at 6-May-2019 19:28 by Biagio Palumbo
    13-Jun-2019 15:35 Statistical engineering challenges from big data in industry
    Statistical Engineering is based on a sequential learning strategy that aims to provide great value to industrial companies through the development and execution of ad-hoc tactics and final solutions. Achieving this goal in industry requires the ability to merge the proper blend of subject matter skills and the smart application of the available tools together.
    The talk will focus on presenting scenarios where the authors have witnessed the virtuous integration of technological skills, traditional statistical methods and modern analytics to face challenges from complex industrial contexts.
    In particular, the first scenario regards the CO2 emission prediction and monitoring problem in the shipping industry that was triggered by the recent EU regulation 2015/757 which came into force in January 2018. Shipping operators have installed on-board large-scale multi-sensor systems for daily monitoring, reporting and verification (MRV) of CO2 emissions for their individual ship. The resulting huge amount of navigation data available has opened up new challenges to translate the statistical engineering paradigm into this context.
    The second scenario regards the manufacturing industry, in the view of the current revolution and modern sensing technologies that naturally benefit from the cross-fertilization of traditional statistical quality control and experiment analysis procedures with modern analytics and data visualization tools. In particular, we aim to provide value to the context of modern additive manufacturing processes, according to the statistical engineering paradigm.
    The third scenario concerns the high-speed rail industry that is acting as a catalyst for industry growth. To improve the performance of critical systems, railway companies have been collecting and storing data to monitor rolling stock, understand his behaviour, and improve his reliability, maintenance and safety. In particular, we introduce a statistical approach for detecting anomalies in critical systems and to allow operators to take prompt actions to reduce, or even avoid, the occurrence of failures.
  • Individual-level patterns of perceived stress throughout the migraine cycle: A longitudinal cohort study using daily prospective data

    Authors: Marina Vives-Mestres (Universitat de Girona), Serena L. Orr (Pediatrics, University of Calgary), Stephen Donoghue (Curelator, Inc.), Kenneth Shulman (Curelator, Inc.), Alec Mian (Curelator, Inc.)
    Primary area of focus / application: Modelling
    Keywords: migraine, personalized medicine, cluster, stress, headache, likert response variable
    Submitted at 17-May-2019 17:58 by Marina Vives-Mestres
    13-Jun-2019 17:20 Individual-level patterns of perceived stress throughout the migraine cycle: A longitudinal cohort study using daily prospective data
    Objective: To determine individual patterns of perceived stress across stages of the migraine cycle.
    Methods: Individuals with migraine registered to use the mobile headache diary N1-Headache® and completed 90 days of daily data
    entry, including perceived stress, rated on a 0-10 scale. Days were categorized into phases: P1 = prodromal (2 days prior to the first
    day with migraine), P2 = migraine (migraine days per International Classification of Headache Disorders-3 definition), P3 =
    postdromal (2 days following the last migraine day with migraine), P0 = interictal days (other days). Individuals with at least 5 days
    in each phase were eligible and data from their first 90 days were used. The odds of stress were modeled with a multinomial
    regression model using sex, age and phase as covariates. A two-step cluster analysis (hierarchical and a k-means) was used to
    determine the number of patterns of stress variation.
    Results: In 730 participants (n= 730), the mean perceived stress rating was 3.4 (standard deviation = + 2.4) and the median was 3.0
    (interquartile range = 3.0). The odds of high perceived stress scores increased in P2 and to a lesser extent in P1 relative to P0
    (p<0.0001), in females relative to males, and decreased with age (p<0.05). Cluster analysis uncovered 6 dominant patterns of stress
    variation. Although P2 had the highest odds of elevated perceived stress scores in the regression model, results of the cluster analysis
    indicate that this is only true for 3 clusters of participants (cluster 1: n=205, cluster 3: n=78 and cluster 6: n=156). Other interesting
    and distinct patterns were seen in clusters 2 (n=79), 4 (n=136) and 5 (n=75).
    Conclusion: Although on an aggregate level perceived stress peaks during the pain phase, in individuals there appear to be 6 distinct
    dominant patterns of stress variation across the migraine cycle. A better understanding of how stress and other related factors vary
    across the migraine cycle in individuals may allow for insights into disease biology and facilitate targeted individualized treatment
    plans in the future.