Scholarlink

<p>We propose a semi-parametric two-component model for the analysis of mixed case interval censored (MCIC) data with a cured subgroup. Such data occurs when the time to an event of interest is only known to belong to an interval obtained from a sequence of, say, <i>k</i> random examination time points with <i>k</i> representing an integer. Furthermore, there is a proportion of subjects who would never be susceptible to the event. The first component of the proposed model describes the probability of cure, and it replaces the traditional generalized linear model with a more flexible support vector machine (SVM)-based approach capable of capturing complex covariate effects. The second component of the proposed model describes the survival distribution of the uncured and is modeled using a Cox proportional hazards structure to preserve the easy interpretation of covariate effects. To the best of our knowledge, this is the first work that employs a machine learning algorithm to analyze MCIC data in the presence of a cured subgroup. To estimate the model parameters, we develop an expectation maximization algorithm. A detailed simulation study demonstrates the superiority of the proposed SVM-based model. Finally, we analyze NASA’s Hypobaric Decompression Sickness Data using the proposed approach.</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

A Support vector machine-based mixture cure model for mixed case interval censored data

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月16日

查看全文
  • Suvra Pal,
  • Wisdom Aselisewine

摘要

We propose a semi-parametric two-component model for the analysis of mixed case interval censored (MCIC) data with a cured subgroup. Such data occurs when the time to an event of interest is only known to belong to an interval obtained from a sequence of, say, k random examination time points with k representing an integer. Furthermore, there is a proportion of subjects who would never be susceptible to the event. The first component of the proposed model describes the probability of cure, and it replaces the traditional generalized linear model with a more flexible support vector machine (SVM)-based approach capable of capturing complex covariate effects. The second component of the proposed model describes the survival distribution of the uncured and is modeled using a Cox proportional hazards structure to preserve the easy interpretation of covariate effects. To the best of our knowledge, this is the first work that employs a machine learning algorithm to analyze MCIC data in the presence of a cured subgroup. To estimate the model parameters, we develop an expectation maximization algorithm. A detailed simulation study demonstrates the superiority of the proposed SVM-based model. Finally, we analyze NASA’s Hypobaric Decompression Sickness Data using the proposed approach.