The use of sparse inverse covariance estimation for relationship detection and hypothesis generation in strategic management

• 成果类型: Article
• 署名作者: Li, Mei; Lin, Ying; Huang, Shuai; Crossland, Craig
• 署名单位: University of Notre Dame; University of Washington; University of Washington Seattle
• 刊物名称: STRATEGIC MANAGEMENT JOURNAL
• ISSN/ISSBN: 0143-2095
• DOI: 10.1002/smj.2450
• 发表日期: 2016
• 页码: 86-97
• 关键词: sparse inverse covariance estimation; Gaussian graphical models; hypothesis generation; relationship detection in high-dimension-low-observation data; exploratory studies

摘要

Research summary: This paper uses Sparse Inverse Covariance Estimation (SICE) to advance strategic management research, focusing on an application of exploratory SICE techniques to generate novel, testable hypotheses. We demonstrate how SICE can identify intrinsic relationships among variables, especially within large, high-dimension-low-observation datasets. We also discuss the strengths and limitations of SICE, as well as the most appropriate uses of these techniques. We conclude with a detailed illustration of SICE analysis using the High Performance Manufacturing dataset.Managerial summary: Most academic research in strategic management is based on formal hypotheses, which are testable statements of the relationships linking two or more variables. Researchers currently use several approaches to generate hypotheses, but all have different shortcomings. In our study, we demonstrate how researchers can use a set of quantitative techniques known as Sparse Inverse Covariance Estimation (SICE) to generate new hypotheses from large datasets. Copyright (c) 2015 John Wiley & Sons, Ltd.