ENBIS: European Network for Business and Industrial Statistics

My abstracts

The following abstracts have been accepted for this event:

• Integer-Valued Autoregressive Models for Counts Showing Underdispersion
  
  Authors: Christian Weiß (Department of Mathematics and Statistics, Helmut Schmidt University Hamburg)
  Primary area of focus / application: Modelling
  Keywords: Count data time series, Good distribution, INAR(1) model, Lerch distribution, Underdispersion, Weighted Poisson distribution
  

  Submitted at 23-Jan-2014 20:31 by Christian Weiß
  Accepted
  

  23-Sep-2014 09:00 Integer-Valued Autoregressive Models for Counts Showing Underdispersion
  The Poisson distribution is a simple and popular model for count-data random variables, but it suffers from the equidispersion requirement, which is often not met in practice. While models for overdispersed counts have been discussed intensively in the literature, the opposite phenomenon, underdispersion, has received only little attention, especially in a time series context. We start with a detailed survey of distribution models allowing for underdispersion, discuss their properties and highlight possible disadvantages. After having identified two model families with attractive properties as well as only two model parameters, we combine these models with the INAR(1) model (integer-valued autoregressive), which is particularly well suited to obtain autocorrelated counts with underdispersion. Properties of the resulting stationary INAR(1) models and approaches for parameter estimation are considered, as well as possible extensions to higher order autoregressions. Three real-data examples illustrate the application of the models in practice.
  
  Weiß, C.H.: Integer-valued Autoregressive Models for Counts Showing Underdispersion.
  Journal of Applied Statistics 40(9), pp. 1931-1948, 2013.
• Perceiving the World through Statistics: Some Dimensional Thoughts
  
  Authors: Michel Lutz (Lutz Industries Research & Consulting, OCTO Technology), Rodolphe Le Riche (CNRS, ENSMSE)
  Primary area of focus / application: Education & Thinking
  Keywords: Epistemology, Dimensions reduction, Statistical thinking, Knowledge generation
  

  Submitted at 24-Jan-2014 15:24 by Michel Lutz
  Accepted
  

  23-Sep-2014 15:20 Perceiving the World through Statistics: Some Dimensional Thoughts
  Using Abbott’s Flatland novel as a metaphor, we will show that our understanding of the reality depends on our capacity to perceive the multiple dimensions of the world. Perceived reality is different for two-dimensional beings than for three-dimensional beings. We propose to develop an analogy with statistics: couldn’t it be said that statistics is the art of trying to understand a high-dimensional world, by projecting it into reduced and intelligible subspaces? Consequently, different data collection strategies, different statistical methods, or different usages, may bring to different understandings of a single world. None of them are true, none of them are false. They are just different reduced perceptions of a complex and infinite-dimensional world. Their discrepancies may lead to complementary data gathering or statistical analysis.
• Statistics in Context: Grounding Quantitative Decision-Aid in Business Needs
  
  Authors: Michel Lutz (Lutz Industries Research & Consulting, OCTO Technology), Xavier Boucher (Ecole des Mines de Saint-Etienne)
  Primary area of focus / application: Business
  Keywords: Decision-aid process, Contextualization, Quantitative and qualitative modelling, Semiconductor industry, Applied statistics
  

  Submitted at 25-Jan-2014 19:24 by Michel Lutz
  Accepted
  

  23-Sep-2014 15:00 Statistics in Context: Grounding Quantitative Decision-Aid in Business Needs
  We propose to present some insights coming from our past research on a semiconductor wafer production plants. The general objective was to implement statistical methods, to improve business decisions. However, we rapidly discovered that introducing statistics in practice is not straightforward when people are not used to it. Consequently, we developed a comprehensive decision-aid process, based on an operation research framework proposed by A. Tsoukiàs. The process is divided in two main stages. Firstly, the context of the decision and the decision problem are formalized. We used a qualitative case-study methodology to build this formalization. The second stage aims at providing quantitative answers to the decision problem. In this perspective, we first built and validated a statistical model, before intergrating it in a specific user-oriented decision-aid analysis. As a result, the model finally provide a quantitative output that fits exactly to the business decision-maker’s needs. This shows that the question of statistical modelling is not sufficient when developing statistical approaches in business contexts: it is necessary to formalize and analyze the application context, to refer to expertise for specific validation steps, and to integrate the models obtained within user-oriented decision-aid analyses or tools.
• Quickly Communicating DoE to Subject Matter Experts
  
  Authors: Matthew Barsalou (BorgWarner Turbo Systems Engineering GmbH)
  Primary area of focus / application: Design and analysis of experiments
  Keywords: Design of Experiments, Fractional factorial, Terminology, Basics
  

  Submitted at 27-Jan-2014 12:39 by Matthew Barsalou
  Accepted (view paper)
  

  23-Sep-2014 16:20 Quickly Communicating DoE to Subject Matter Experts
  To perform a Design of Experiments (DoE) on a complex process technical support from the process experts is often needed. The process experts should be there to help identify which factors may be relevant and should be explored and which do not need to be evaluated. Expert advice is especially important in performing factional factor DoEs where not all potential levels would be considered.
  
  There may be problems when discussing a DoE with a subject matter expert who lacks a basic understanding of DoE. A subject matter expert may not need to know the word “response variable”; however, if they don’t understand the concept, they can’t help to identify the correct response variable for a DoE.
  
  This talk will present a simple concept for teaching engineers and other subject matter experts the basic concepts needed for them to contribute to a DoE. It will include explanations for factors, response variable, levels, experimental runs, blocking, randomization and resolution that are easy to communicate to those without a background in DoE and will also explain how to give a one minute DoE “elevator speech” using a paper helicopter.
• Mosaic Plots for Visualizing Confounding Properties of Factorial Designs
  
  Authors: Ulrike Grömping (Beuth Hochschule für Technik Berlin)
  Primary area of focus / application: Design and analysis of experiments
  Keywords: Factorial experiment, Confounding, Design of Experiments, Mosaic plot
  

  Submitted at 27-Jan-2014 18:24 by Ulrike Grömping
  Accepted (view paper)
  

  23-Sep-2014 09:40 Mosaic Plots for Visualizing Confounding Properties of Factorial Designs
  Factorial experiments are widely used in industrial experimentation and other fields. Whenever a factorial experiment is not designed as a full factorial but as a – regular or non-regular – fraction thereof, choice between competing designs and interpretation of experimental results should take into consideration, how the experimental plan will confound experimental effects. According to the effect hierarchy principle, the lowest degree confounding is considered most severe, i.e. is particularly important to assess.
  
  Mosaic plots are a tool for visualizing the structure of multidimensional frequency tables. They are usually attributed to Hartigan and Kleiner (1981 and 1984) and have met with increased attention lately (e.g. Hofmann 2003, Wickham and Hofmann 2011). A frequency table of several factors in a factorial design captures the confounding structure among these factors. This talk proposes to use mosaic plots of such frequency tables for visualizing the degree of confounding. Mosaic plots are particularly useful for design and analysis of orthogonal main effects plans, which are usually of resolution III only.
  
  The plots are available in open source software: the R package DoE.base (Grömping 2013) creates them, based on the R-package vcd (Meyer and Hornik 2006).
  
  References
  Grömping, U. (2013). DoE.base: Full factorials, orthogonal arrays and base utilities for DoE packages. R package version 0.25-2. In R Core Team (2013). R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria.
  
  Hartigan, J. A., and Kleiner, B. (1981). Mosaics for contingency tables. In W. F. Eddy (Ed.), Computer Science and Statistics: Proceedings of the 13th Symposium on the Interface. New York: Springer-Verlag.
  
  Hartigan, J. A., and Kleiner, B. (1984). A mosaic of television ratings. The American Statistician 38, 32-35.
  
  Hofmann, H. (2003). Constructing and Reading Mosaicplots. Computational Statistics and Data Analysis 43, 565–580.
  
  Meyer, D., Zeileis, A. and Hornik, K. (2006). The Strucplot Framework: Visualizing Multi-way Contingency Tables with vcd. Journal of Statistical Software 17, Issue 3, 1:48.
  
  Wickham, H. and Hofman, H. (2011). Product Plots. IEEE Transactions on Visualization and Computational Graphics 17, 2223-2230.
• Half Normal Effects Plots in the Presence of a Few Error Degrees of Freedom
  
  Authors: Ulrike Grömping (Beuth Hochschule für Technik Berlin)
  Primary area of focus / application: Design and analysis of experiments
  Keywords: Half normal effects plot, Lenth’s method, (Almost) unreplicated experiment, Orthogonalization of error space, Augmented half normal effects plot
  

  Submitted at 27-Jan-2014 18:33 by Ulrike Grömping
  Accepted (view paper)
  

  23-Sep-2014 10:55 Half Normal Effects Plots in the Presence of a Few Error Degrees of Freedom
  Fractional factorial 2-level experiments are often conducted without any error degrees of freedom. In such cases, a half-normal effects plot – also called Daniel plot according to its inventor (Daniel 1959, 1976) – can be used for assessing effect significance. Half-normal effects plots are often accompanied by a numerical method for assessing effect significance, most prominently Lenth’s method (Lenth 1989). There are, however, also situations for which a few error degrees of freedom are available, for example from a replicated center point run. For such cases, besides the obvious possibilities of either ignoring the few replicates (i.e. using half-normal effects plot and Lenth’s method, as if they were not there) or using analysis of variance with the replicates for estimating the error variance, several further proposals for assessing effect significance exist (Larntz and Whitcomb 1998, Edwards and Mee 2008, JMP software). This talk compares the published methods, proposes an additional one (that might be very close to what JMP does) and advocates the use of an augmented half-normal effects plot that shows error points along with the effects. It is argued that such a plot can even be useful in a fully-replicated experiment for assessing whether the replication process was of adequate quality. The method is available in the R package DoE.base (Grömping 2013).
  
  References
  
  Daniel, C. (1959). Use of Half-normal effects plots in Interpreting Two Level Experiments. Technometrics 1, 311–340.
  
  Daniel, C. (1976). Application of Statistics to Industrial Experimentation. Wiley, New York.
  
  Edwards, D. J. and Mee, R. W. (2008). Empirically Determined p-Values for Lenth t-Statistics. Journal of Quality Technology 40, 368–380.
  
  Grömping, U. (2013). DoE.base: Full factorials, orthogonal arrays and base utilities for DoE packages. R package version 0.25-2. In R Core Team (2013). R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria.
  
  Larntz, K. and Whitcomb, P. (1998). Use of replication in almost unreplicated factorials. Manuscript of a presentation given at the 42nd ASQ Fall Technical conference in Corning, New York. Downloaded 4/26/2013 at http://www.statease.com/pubs/use-of-rep.pdf.
  
  Lenth, R.V. (1989). Quick and easy analysis of unreplicated factorials. Technometrics 31, 469–473.