ENBIS-17 in Naples

9 – 14 September 2017; Naples (Italy) Abstract submission: 21 November 2016 – 10 May 2017

My abstracts

 

The following abstracts have been accepted for this event:

  • A Research Framework Based on Modeling and Simulation for Expected Net Benefits and Making Decisions on Willingness to Pay per Health Gain when Conducting Randomized Clinical Trials

    Authors: Ismail Abbas (Universitat Politecnica de Catalunya)
    Primary area of focus / application:
    Keywords: Clinical trials, Expected net benefits, Statistics, Simulation
    Submitted at 16-Apr-2017 23:59 by Ismail Abbas
    Accepted (view paper)
    11-Sep-2017 16:20 A Research Framework Based on Modeling and Simulation for Expected Net Benefits and Making Decisions on Willingness to Pay per Health Gain when Conducting Randomized Clinical Trials
    Introduction: The expected net benefits of clinical trials depend on willingness to pay per health gain, and the aim is to show what amount of willingness to pay should be attached to health benefits when testing the hypothesis that there is or not a statistical evidence of ENB-expected net benefit.

    Methods: The framework considers two-stages modeling and simulation with prior information of a clinical trial. The first-stage calculates sample size and power for testing a primary hypothesis of the trial. Giving the resulting power and sample size, the hypothesis of ENB is tested to make decision on power and willingness to pay per health gain.

    Results: The framework is populated with a clinical trial data on benefits including efficacy, effectiveness and health care cost, combined with an assumption on correlated distributions, market exclusivity period and trial cost. The trial compares the magnetic resonance imaging-MRI with computerized axial tomography scanner-CT interventions in the diagnostic of acute ischemic stroke, assessing that CT is the dominant intervention by means of cost-effectiveness analysis. The ENB is €26M for 0.8 (80%) of power, at which €41000 should be attached per unit of health benefits. Sensitivity-analyses results and optimal ENB for a given willingness to pay will also be presented.

    Discussion and conclusions: Simulation modeling is useful when making decision on a reasonable willingness to pay per health gain for randomized clinical trials that incorporate expected net benefits analysis. Nonetheless, the framework might be considered as a base for further developments and applications.
  • Optimal Design of Xbar Control Charts under Bathtub-shaped Shock Models with Application of Total Time on Test Plot

    Authors: Shabnam Fani (Allameh Tabataba'i University), Mojtaba A. Pasha (Allameh Tabataba'i University), Mohammad B. Moghadam (Allameh Tabataba'i University)
    Primary area of focus / application: Quality
    Secondary area of focus / application: Reliability
    Keywords: Control chart design, Process failure mechanism, Bathtub-shaped hazard rate, Total time on test plot, Integrated hazard over sampling intervals
    Submitted at 17-Apr-2017 09:44 by Shabnam Fani
    Accepted
    13-Sep-2017 10:10 Optimal Design of Xbar Control Charts under Bathtub-shaped Shock Models with Application of Total Time on Test Plot
    Some recent research in statistical quality control has paid attention to the economic and economic statistical design of X bar-control charts for monitoring the mean variation of a production process. The majority of these studies assumed that the process failure mechanism or the shock model obeys a lifetime distribution with monotone hazard rate such as exponential, Weibull, and Gamma. However, virtually sparse research efforts have been investigated non-monotone failure mechanism such as bathtub or U-shaped. In order to choose a more precise lifetime distribution, we propose to apply the total time on test plot as a great tool for identification of the shape of failure data set. Based on the result of this plot, an appropriate distribution will be picked out whose unknown parameters can be estimated by the maximum likelihood method. Finally, the optimal parameters of both the economic and economic statistical design of Xbar-control charts for systems with bathtub-shaped shock models can be determined by adapting the non-uniform generalized cost model of Rahim and Banerjee (1993). A numerical study is presented to illustrate the proposed approach for a exact failure time data set.
  • Bayesian Adaptive Planning in Type II Step-Stress Accelerated Life Tests with Multiple Steps

    Authors: Masakatsu Onoda (Tokyo University of Science), Seiichi Yasui (Tokyo University of Science)
    Primary area of focus / application: Reliability
    Keywords: Bayesian estimation, Inverse power law, Bayesian decision making, Weibull distribution, Asymmetry absolute loss
    Submitted at 17-Apr-2017 13:55 by Seiichi Yasui
    Accepted (view paper)
    12-Sep-2017 10:10 Bayesian Adaptive Planning in Type II Step-Stress Accelerated Life Tests with Multiple Steps
    One of the accelerated life tests to economically estimate the lifetime distribution of the product is a step-stress test in which the stress levels are changed in stages within the test. In general, the stress levels are monotonically increased or decreased, and the amount of stress for each step and timing of the change are predetermined before the test. Therefore, the test plan needs to be set up with information on the test items. Above all it is important to set appropriate levels of stress for each step. However, if the information on the test items is insufficient, many test items may instantly fail when raising the stress level, or no failure is observed in the step at all, which leads to poor estimation accuracy.
    We propose type II step-stress accelerated life tests with multiple steps, in which the stress level for each step is adaptively determined by Bayesian estimation of the lifetime distribution based on failure times previously obtained during the life test. The adaptive determination of stress levels makes it possible to control the probability that some items to be tested in a step fail within minimum unit of observational time after changing stress levels.
    In this paper, it is assumed that lifetimes are distributed according to the Weibull distribution. We deal with some asymmetric absolute loss functions in Bayesian decision making to control the probability of multiple failures at changing stress levels. The estimation accuracy of parameters in the inverse power law is evaluated through Monte Carlo methods.
  • Multi-Factor Online Testing

    Authors: David Steinberg (Tel Aviv University), Tamar Haizler (Tel Aviv University)
    Primary area of focus / application: Design and analysis of experiments
    Secondary area of focus / application: Business
    Keywords: DoE, A/B testing, Click-through rate, Factorial experiments, Thompson sampling, Bandit methods
    Submitted at 17-Apr-2017 15:40 by David Steinberg
    Accepted (view paper)
    12-Sep-2017 09:40 Multi-Factor Online Testing
    Use of designed experiments has become a major tool in web site design, helping companies market products and services more successfully. The experiments are used to compare possible changes to web pages with respect to outcomes like “click-through rate”, “download rate” or “total sales”. Most of these experiments compare two formats for the web page, hence the common term "A/B" testing. This talk will explore methods for the design and analysis of online experiments that look simultaneously at a number of factors. Important features are to balance the need to explore and to exploit and to take advantage of the large traffic volume characteristic of most internet sites.
  • A Generalized Maximum Entropy Approach for Analyzing Unreplicated Factorial Designs

    Authors: Inara Francoyse de S. Pereira (Universidade Federal do Rio Grande do Norte), Carla A. Vivacqua (Universidade Federal do Rio Grande do Norte), Andre Luis S. de Pinho (Universidade Federal do Rio Grande do Norte)
    Primary area of focus / application: Other: DOE and statistical process monitoring in South America
    Secondary area of focus / application: Design and analysis of experiments
    Keywords: Cost of experimentation, Design of Experiments, Fractional factorial design, Normal probability plot, Ordinary least squares, Replication, Uncertainty
    Submitted at 17-Apr-2017 16:16 by Carla Vivacqua
    Accepted
    11-Sep-2017 18:10 A Generalized Maximum Entropy Approach for Analyzing Unreplicated Factorial Designs
    Studies to investigate the effects of various factors on processes are common. Factorial experiments are widely used in these cases. There are situations where replication becomes impractical due to lack of resources or time. One popular method for analyzing unreplicated factorial designs is based on normal probability plots. Disadvantages of this approach include subjectivity and difficulty of interpretation, especially when the number of treatments is small and the magnitude of the effect is moderate. This presentation aims to discuss an alternative way to evaluate factorial designs. The proposed approach is based on generalized maximum entropy. First, we present the construction of the method and show how to use it to identify active effects. Next, a set of case studies are illustrated and comparisons between both approaches are made. Then, a simulation study indicates that the generalized maximum entropy approach makes it easier to identify active effects.
  • Anomaly Detection in Maritime Data Streams

    Authors: Ingrid Kristine Glad (University of Oslo), Andreas Brandsæter (University of Oslo/DNV-GL), Martin Tveten (University of Oslo)
    Primary area of focus / application: Other: Novel methods for industrial data streams
    Keywords: Sequential testing, Multiple sensors, Sparsity, Anomaly detection
    Submitted at 17-Apr-2017 23:07 by Ingrid Kristine Glad
    Accepted
    12-Sep-2017 16:00 Anomaly Detection in Maritime Data Streams
    In the shipping industry, there is an increasing interest in monitoring ship operations through sensor data continuously streamed from vessels to shore. Data streams represent very diverse aspects of the ship´s state and performance, while the ships operate in several modes and under various climatic conditions.
    In order to be fed into (sequential) tests for anomaly detection, such data streams are first taken through a data driven, nonparametric method for signal reconstruction known as Auto Associative Kernel Regression (AAKR). Streams of residuals (observed - reconstructed) are then monitored successively. We compare the classical univariate Sequential Probability Ratio Test (SPRT) with more recent extensions to multivariate (and possibly high dimensional) procedures for sequential detection of changes in mean and/or variance and covariance. These sequential tests are also compared to other change point detection procedures, with the aim of detecting changes as soon as possible, while controlling the amount of false alarms.