2016 INFORMS Annual Meeting Program

TD03

INFORMS Nashville – 2016

2 - Graph Mining For Alzheimer Disease Fei Gao, Arizona State University, Tempe, AZ, 85281, United States, fgao16@asu.edu, Teresa Wu

3 - Vertical Probabilistic Selling Under Competition: The Role Of Consumer Anticipated Regret Dongyuan Zhan, University College London, London, WC1E 6BT, United Kingdom, d.zhan@ucl.ac.uk, Yong Chao, Lin Liu We study probabilistic selling with vertically differentiated products when firms compete and consumers anticipate the potential post-purchase regret raised by obtaining the inferior products. Intuitively, anticipated regret hurts the attractiveness of probabilistic selling. However, we find that probabilistic selling can be more profitable, and more likely to arise with anticipated regret than without it. That is due to the “reverse quality discrimination” (perceived quality of the random product is non-increasing in consumer type), which increases the perceived differentiation at the competition margin, and maintains the random products attractive to the infra-marginal consumers. TD04 101D-MCC Joint Session QSR/ENRE: Data-driven Modeling and Analytics in Wind Power Systems Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Arash Pourhabib, Oklahoma State University, 322 Engineering North, Stillwater, OK, 74078, United States, arash.pourhabib@okstate.edu Co-Chair: Eunshin Byon, University of Michigan, College Station, MI, United States, ebyon@umich.edu 1 - Extreme Loads Analysis: Extrapolation And Importance Sampling Peter Graf, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, CO, 80401, United States, Peter.Graf@nrel.gov Assessing wind turbine extreme loads requires estimating tails of probability distributions to construct “exceedance plots” of probability versus peak loads in a 10 minute simulation corresponding to a once-in-50-years event. The IEC standard contains a prescription for how to estimate these loads. Many find it unsatisfying because it relies on extrapolation to achieve the 10E-8 level. Alternative methods based on more strategic sampling of conditions are promising because they may allow for direct statistical realization of the extreme loads. This paper compares the existing IEC approach to one based on Importance Sampling. 2 - Space-time Modeling Of Asymmetric Local Wind Fields Ahmed Aziz Ezzat, Texas A&M University, College Station, TX, United States, aa.ezzat@tamu.edu, Mikyoung Jun, Yu Ding Local wind fields refer to the wind dynamics in a space-time domain composed of a dense grid of locations with close space-time proximity. A typical application is modeling wind stream behavior using measurements at wind turbines on a wind farm. Existing literature tends to overlook space-time interaction by imposing separable, symmetric models. Our analysis suggests that local wind dynamics are asymmetric in nature, and this asymmetry pattern is dynamically changing due to alternation of dominant winds. Modeling such physical phenomenon can have a vital impact on our understanding of local wind dynamics, enabling better forecasts and robust control strategies in wind energy applications. 3 - Data-driven Stochastic Transmission Expansion Planning Ali Bagheri, Oklahoma State University, ali.bagheri@okstate.edu Due to the significant improvements of power generation technologies and replacing traditional power plants with renewable ones, the generation portfolio will experience dramatic changes. The uncertainty of renewable energy and their sitting call for economic plans for expanding the transmission capacities. In this study, by learning from the historical data, we first construct a confidence set for the unknown distribution of the uncertain parameters. Then, we develop a two- stage data-driven transmission expansion framework, by considering the worst-case distribution within the constructed confidence set. To tackle the model complexity, we propose a decomposition framework. 4 - Data-driven Approach For Wake Effect Analysis: Generalization To All Wind Directions Mingdi You, University of Michigan, mingdyou@umich.edu, Eunshin Byon, Judy Jin Utility-scale wind farms consist of a large number of turbines. To improve the power generation efficiency of turbines, accurate quantification of power generations of multi-turbines is critical in wind farm design and operations. One challenging issue is that the power outputs of multiple turbines are different because of complex interactions among turbines, known as wake effects. In general, downstream turbines tend to produce less power than upstream turbines. When wind direction changes, such wake correlations among turbines also change. This study proposes a new statistical approach that quantifies the wake effects on power generations under different wind directions.

Apolipoprotein E (APOE) is a gene considered to be highly correlated with the risks of having Alzheimer’s disease (AD). In this study, imaging data (T1 and DWI) of two cohorts of patients: APOE carrier and non-carrier is studied. Brain network were first generated based on which linear regression on graphic features such as clustering coefficient, mean degree versus age was conducted. The results showed there may be differences between the two cohorts. This finding motivated us further to develop a multi-task feature selection method to identify the critical sub-graphic regions that can significantly improve the discriminative power using the new graph features versus age correlation. 3 - Mining Hierarchical Event Labels In Large-scale Eeg Collections When comparing results from two similar EEG studies, researchers must manually map event types in one study to those of the other. To address event mapping and to facilitate large-scale data sharing, we created an event-labeling scheme called Hierarchical Event Descriptors (HED). HED is a common, extensible vocabulary, with more detailed tags appearing lower in the hierarchy. We tagged over 3 million events across 1,127 datasets using the HED tag system. We perform a cross study analysis by investigating ERP/ERSP patterns calculated by averaging over data recording trials extracted by matching HED tags. We evaluate the correlations of these patterns and their relationship to particular tags. 4 - Capturing Dynamics Of Brain Functional Networks Through Data Driven Techniques Laleh Najafizadeh, Rutgers University, laleh.najafizadeh@rutgers.edu The brain is a highly complex dynamic system in which neuronal connections are continuously formed and dissolved at multiple temporal scales. A challenging problem in the field of neuroscience has been to find reliable techniques that can describe such inherently dynamic properties of brain function. One promising approach to investigate brain’s functional architecture is to study its function at the network level within the context of functional connectivity. Utilizing this approach, here, we present data driven frameworks to examine the dynamic nature of neuronal activity, during the execution of tasks. Kay Robbins, University of Texas at San Antonio, Kay.Robbins@utsa.edu, Nima Bigdely-Shamlo Chair: Tingliang Huang, Carroll School of Management, Boston College, Chestnut Hill, MA, United States, tingliang.huang@bc.edu 1 - Selling Through Priceline? On The Impact Of Name-your-own-price In Competitive Market Xiao Huang, John Molson School of Business, Concordia University, xiao.huang@concordia.ca, Greys Sosic, Gregory E Kersten We study how competitive sellers with substitutable goods may sell their products (1) as regular goods, through a direct channel at posted prices, and/or (2) as opaque goods, through a 3rd-party NYOP channel. We establish a stylized framework with two competing sellers, an intermediary NYOP firm, and a sequence of customers. We characterize customers’ optimal purchasing/bidding decisions and sellers’ dynamic pricing equilibrium, and then illustrate the impact of inventory and time on prices, profits and channel strategies via numerical studies. Interestingly, although competing sellers seldom benefit from the existence of NYOP, it is possible that some seller(s) may adopt it in equilibrium. 2 - Opaque Selling And Last-minute Selling: Revenue Management In Vertically Differentiated Markets Hang Ren, School of Management, University College London, London, United Kingdom, hang.ren.13@ucl.ac.uk, Tingliang Huang Firms in many industries often reduce the price of products/service at the end of the selling season to dispose of unsold inventory/capacity. This last-minute selling induces consumers to wait for sales and thus lowers the regular price. To overcome the problem, many firms switch to opaque selling, i.e., mixing different types of leftovers and sell them as one type of product. We study the performance of last-minute selling and opaque selling in vertically differentiated markets, and find that opaque selling is less efficient in cleaning up leftovers, and the firm may switch to last-minute selling when high demand becomes more likely. Both results are contrary to the horizontal differentiation case. TD03 101C-MCC Recent Developments in Opaque and Probabilistic Selling Invited: Business Model Innovation Invited Session

334