---
格式版本: 2
标题: "Partial Identification Learning with Categorical Treatments for Individualized Treatment Rules"
原文链接: "https://arxiv.org/abs/2608.19853"
发布日期: "2026-08-20"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Thu, 20 Aug 2026 10:00:00 UTC (166 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-21T16:18:31+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-21T16:17:02+08:00"
入库时间: "2026-08-21T08:18:32.145Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Oracle&searchtype=all"
匹配关键词:
  []
相关厂家:
  - "Oracle"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 0
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该页面为arXiv统计学方法论论文，研究分类治疗规则的偏识别学习，与超节点、AI Rack、机柜级AI基础设施等主题完全无关，无任何相关技术或商业信息。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-21T16:18:40+08:00"
AI主题相关性: 0
AI来源权威性: 0
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "该文提出一个面向类别型治疗、结局和工具变量的个体化治疗规则部分识别学习框架，利用因果界限刻画最优治疗决策，并将现有二元治疗设定推广到更广的因果结构。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T02:12:10.885Z"
采集批次: "2026年8月21日13点55分53秒"
采集批次ID: "20260821-135553-743"
去重键: "https://arxiv.org/abs/2608.19853"
---

## Statistics > Methodology

## Title:Partial Identification Learning with Categorical Treatments for Individualized Treatment Rules

[View PDF](https://arxiv.org/pdf/2608.19853) [HTML (experimental)](https://arxiv.org/html/2608.19853v1)

> Abstract:We develop a partial identification learning framework for individualized treatment rules (ITRs) with categorical treatments, outcomes, and instrumental variables. Rather than relying on strong causal assumptions required for point identification, our framework leverages causal bounds to characterize the optimal treatment decision. Existing methods for ITR optimization under partial identification are largely restricted to binary treatment settings and the bounds derived by Balke and Pearl under the canonical instrumental variable design. We extend this framework to accommodate a broader class of causal structures as well as scenarios with categorical treatment, outcome, and instrumental variables. We introduce a generalized minimax loss criterion for treatment selection from among more than two options, which minimizes the maximum possible difference between the chosen and the optimal treatment based on partial identification bounds. To construct the ITR, we use a symmetric embedding strategy that maps discrete treatments to the vertices of a regular simplex, avoiding the geometric inconsistencies of standard one-vs-rest approaches. We derive a differentiable, weighted surrogate risk function and show that optimizing it solves the original problem. Furthermore, we provide finite sample convergence rates via an oracle inequality under general regularity conditions, which we show are satisfied by a kernel based implementation. Numerical experiments demonstrate that the framework yields ITRs significantly closer to the oracle ITR compared to existing alternatives in settings with unmeasured confounding.

| Subjects: | Methodology (stat.ME) |
| --- | --- |
| Cite as: | [arXiv:2608.19853](https://arxiv.org/abs/2608.19853) \[stat.ME\] |
|  | (or [arXiv:2608.19853v1](https://arxiv.org/abs/2608.19853v1) \[stat.ME\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.19853](https://doi.org/10.48550/arXiv.2608.19853) |

## Submission history

From: Johannes Hruza \[[view email](https://arxiv.org/show-email/ed2083d1/2608.19853)\]  
**\[v1\]** Thu, 20 Aug 2026 10:00:00 UTC (166 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.19853) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))