INFORMS 2021 Program Book

INFORMS Anaheim 2021

SD18

3 - Estimating Direct Causal Effect under Hierarchical Interference On Networks In Observational Studies Fei Fang, Duke University, Chapel Towers Durham, NC, 27705- 4547, United States, Alexandre Belloni, Alexander Volfovsky We study causal estimators for direct treatment effect under interference given network and treatment assignment. Our estimators are constructed by hierarchical trees whose nodes represent distinguished pattern graphs, under interference associated with treated neighborhood graphs. Applying our pruning procedure to these trees, oracle inequalities and corresponding adaptive rates are established to the estimators. Our results are generic to different hierarchical structures of interference and data generating processes. A theoretical example whose explicit rate of convergence is investigated when the graph is generated by stochastic block model. Finally, we examine the empirical performance of our estimators by simulations and real datasets. SD13 CC Room 201A In Person: Network Optimization and its Applications General Session Chair: Illya V Hicks, Rice University, Houston, TX, 77005-1827, United States Co-Chair: Samuel Kroger, Rice University 1 - MIP Formulations for Solving the Maximum Anchored K-core Problem Samuel Kroger, Rice University, Houston, TX, United States, Hamidreza Validi, Illya V. Hicks In this talk, we investigate two mixed integer programming (MIP) formulations for the maximum anchored k-core problem. We compare the MIP formulations analytically and computationally. Furthermore, we propose valid inequalities and fixing procedures to improve their computational performance. Finally, we conduct an extensive set of experiments to evaluate the performance of the MIP formulations. Joint Work with Hamidreza Validi and Illya V. Hicks 2 - An Improved Approximation for Maximum K-dependent Set on Bipartite Graphs Sergiy Butenko, Texas A&M University, College Station, TX, 77841-3131, United States, S. Mohammad Hosseinian We present a (1+k/(k+2))-approximation algorithm for the Maximum k- Dependent Set problem on bipartite graphs for any positive integer constant k. The algorithm runs in $O(k m \sqrt{n})$ time and improves upon the previously best-known approximation ratio of 1+k/(k+1) established by Kumar et al. [Theoretical Computer Science, 526: 90-96 (2014)]. SD15 CC Room 201C In Person: Empirical Research in Health Care Operations General Session Chair: Jong Myeong Lim, The Wharton School, Philadelphia, PA, 19104, United States 1 - Nonprofit vs. For-profit: Allocation of Beds And Access to Care in U.S. Nursing Homes Yangzi Jiang, Northwestern University, Evanston, IL, 60201, United States, Lauren Xiaoyuan Lu, Jan A. Van Mieghem Motivated by empirical observations of U.S. nursing homes, we formulate a queueing network model to study nursing homes’ bed allocation decisions and the resulting access to care for economically disadvantaged populations. To distinguish nonprofit from for-profit nursing homes, we incorporate altruism into a nonprofit nursing home’s objective function to capture resident welfare including the blocking cost of high-margin Medicare residents and the waiting cost of low-margin Medicaid residents. Our theoretical and empirical findings inform the public that the growth of the for-profit nursing home segment does not necessarily hurt the access to care for the Medicaid population, and surprisingly, under high Medicaid demand, for-profit nursing homes might

since these guidelines were published. In this analysis, we investigated the cost- effectiveness of LC by developing a microsimulation model of the natural history of hepatitis C-cured patients and using the latest published data on LC progression, treatments, and surveillance adherence. We found that semi-annual screening is cost-effective in patients with cirrhosis if LC risk>0.4%. SD16 CC Room 201D In Person: Mining Digital Trace Data of Online Communities General Session Chair: Tianjie Deng, University of Denver, United States 1 - Are Critics Really Unbiased? The Impact of Social Ties on Critics’ Rating Behavior Tianxi Dong, Assistant Professor, Trinity University, San Antonio, TX, United States, Tianjie Deng, Thomás Peña Grounded in the differentiation theory, this study aims to empirically investigate the relationship between social ties and the rating similarities between critics. We collected an extensive data set from Rotten Tomatoes exploring the critics’ social relations in conjunction with their movie-rating behavior. We find that loners (critics who have no connections) give higher ratings than non-loners (critics with at least one connection). What is more, critics tend to give lower ratings when they have more connections. In terms of social tie strengths, critics with strong ties appear to provide similar ratings. These findings raise questions about the reliability of critic ratings as unbiased indicators of quality. Platform stakeholders can adjust their strategies to account for possible review biases resulting from the social interactions among critics. 2 - Preserving History: Archiving Search Query Results for Efforts to preserve web pages have been dramatically increased in the past decade. Cheaper data storage and faster processing have allowed for an increase in the number websites archived and the frequency with which they are archived. However, archiving the results of search queries presents unique challenges not easily addressed by traditional web crawling methods. This research outlines some of the potential challenges and presents some possible solutions for archiving these types of pages. 3 - A Reinforcement Learning Algorithm for Online Personalized Tutor Recommendation Mohamad Kazem Shirani Faradonbeh, University of Georgia, Athens, GA, United States Intelligent computerized education reduces costs of tutoring by learning from the trajectories of the students. We present a data-driven algorithm implemented on an online platform for recommending personalized tutoring to students. To do so, multiple important challenges are addressed. First, the experiments for collecting data need to be diverse for exploring student responses, while at the same time they must focus on the immediate weakness of each student. Moreover, there are many tutoring items, but each student provides an extremely small data. Further challenges as well as employed methods that utilize student backgrounds for combining the data will be discussed. SD18 CC Room 202B In Person: Macro Energy Systems: Energy and Climate General Session Chair: Tyler Ruggles, Research Scientist, Carnegie Science, Stanford, CA, United States 1 - Operational Flexibility of Natural Gas Combined Cycle Power Plant Coupled with Flexible Carbon Capture and Storage Fangwei Cheng, Princeton University Achieving net-zero economy requires affordable low carbon or carbon neutral power systems. Natural gas combined cycle (NGCC) coupled with carbon capture and sequestration (CCS) enables continuous consumption of fossil fuels for power generation with minimal CO2 emissions. In this study, we apply integer clustering and linear relaxation unit commitment (UC) to the subcomponents (e.g., gas turbine, steam turbine, absorber, and regenerator) of NGCC coupled with flexible CCS and compare the results against the conventional binary UC. Our results show integer clustering/linear relaxation UC of NGCC-CCS subcomponents leads to substantial run time reduction with marginal errors. We also study how flexible NGCC-CCS affects the economic, environmental, and generation dispatch profiles under a wide range of carbon price (0-120 $/t) and variable renewable capacity. Future Research Joshua Madden

provide higher access to care than their nonprofit counterparts. 2 - Liver Cancer Screening after Hepatitis C Cure: Updating the Outdated Risk Threshold for Screening

Ali Hjaar, Harvard Medical School, Mass General Hospital, Boston, MA, United States, Jagpreet Chhatwal, Peter Mueller, Gizem Nemutlu, Mary L. Peters, Leigh Anne Dageforde, Fasiha Kanwal The number of hepatitis C-cured patients is rapidly rising but these patients remain at risk of liver cancer (LC). The current screening guidelines recommend semi-annual screening if the patient’s annual risk of LC >1.5%. However, this threshold is considered outdated as LC treatments have improved substantially

35