Scholarlink

<p>In this study, we focus on the online monitoring problem, aiming to swiftly identify changes in parameters within bivariate time series models of counts, where counts may potentially exhibit negative values. To accomplish this objective, we introduce the bivariate signed integer-valued autoregressive (BSINAR) model tailored for the analysis of count time series. After introducing the general BSINAR framework, we specialize the innovations to a signed Poisson mixture, which facilitates likelihood-based estimation. Then, utilizing the score vectors from this model to build monitoring processes, we employ two types of detectors to establish stopping rules. The control limits for these processes are determined asymptotically based on the corresponding limit theorems. To evaluate and compare the performance of our methods, we conduct Monte Carlo simulations and provide a real data analysis for illustrative purposes. All our findings consistently support the effectiveness and validity of the proposed monitoring procedures.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sequential monitoring process for bivariate signed integer-valued autoregressive models

  • 期刊论文

  • 发表日期：

    2025年9月25日

查看全文
  • Sangyeol Lee,
  • Minyoung Jo

摘要

In this study, we focus on the online monitoring problem, aiming to swiftly identify changes in parameters within bivariate time series models of counts, where counts may potentially exhibit negative values. To accomplish this objective, we introduce the bivariate signed integer-valued autoregressive (BSINAR) model tailored for the analysis of count time series. After introducing the general BSINAR framework, we specialize the innovations to a signed Poisson mixture, which facilitates likelihood-based estimation. Then, utilizing the score vectors from this model to build monitoring processes, we employ two types of detectors to establish stopping rules. The control limits for these processes are determined asymptotically based on the corresponding limit theorems. To evaluate and compare the performance of our methods, we conduct Monte Carlo simulations and provide a real data analysis for illustrative purposes. All our findings consistently support the effectiveness and validity of the proposed monitoring procedures.