Technical note: Sufficient operational statistics

• 成果类型: Article
• 署名作者: Jia, Justin; Katok, Elena
• 署名单位: University of Tennessee System; University of Tennessee Knoxville; University of Texas System; University of Texas Dallas
• 刊物名称: PRODUCTION AND OPERATIONS MANAGEMENT
• ISSN/ISSBN: 1059-1478
• DOI: 10.1111/poms.13678
• 发表日期: 2022
• 页码: 2429-2437
• 关键词: data-driven decision making; data reduction; operational statistics; sufficiency

摘要

The decision in a data-driven decision-making problem is generally a high-dimensional function of data. When can the decision be reduced to a single-dimensional function of a statistic? This study addresses this question based on the operational statistics literature. The study introduces the notion of sufficient operational statistics and derives the factorization theorem for identifying such statistics. Further, the study proposes a solution procedure based on the statistics and derives the finite-sample performance bound of the proposed solution.