Kuldeep Kumar’s Contribution to the Discussion of ‘Assumption-Lean Inference for Generalised Linear Model Parameters’ by Vansteelandt and Dukes
Kuldeep Kumar 对《广义线性模型参数的弱假设推断》的讨论AI 解读 ↓
← 返回论文列表与筛选
阅读说明
本页整理公开书目信息，不生成学术质量评分。AI 解读根据题名及所注明的摘要或原文片段生成，未完成全文核验，可能有误；具体假设、证明与结论请核对原文。摘要保留来源文本，未经模型改写。
DOI 首次登记不等于接收或发表日期。部分更正、评论等条目没有独立摘要；缺少可靠材料时保留待补状态。
引用量采用 OpenAlex 单条关联记录，不合计多个记录；关联记录可能包含正式发表版本，不等于此版本单独引用。缺失不视为零，引用指标不代表论文质量。
关注作者按书目中的完整姓名匹配，同名和姓名变体需人工辨别。相关论文先按共享细分方向和具体方法排列，再参考通用方法与发表日期。
分类依据与纠错
AI 根据所提供的论文材料归类：文本讨论一般广义线性模型推断的适用性，未提供所列领域细分的直接研究依据。
反馈分类问题
正在核对关联版本…
代码、数据与补充材料
正在核对对应版本的资源…
关注这篇论文的作者
Kuldeep Kumar 对《广义线性模型参数的弱假设推断》的讨论
- 研究问题
- 这篇讨论关注被评文章在小样本、先验信息、多重共线性及非正态协变量下的适用性；文本主要提出问题，没有给出相应研究答案。
- 方法与思路
- 讨论者借助数据建模与算法建模的区分，评价被评文章结合机器学习和传统统计方法的思路，再针对其模拟和应用提出具体疑问。
- 材料陈述的贡献
- 可确认的贡献是提出小样本偏差、贝叶斯先验、共线性及高维非正态协变量等核查方向；文本未提供新的估计量或验证结论。
- 阅读关注点
- 应将这些疑问与主论文及作者回复对照，核对样本规模、偏差表现和适用条件，区分讨论者的评价、问题与主论文实际证明的结论。
- 阅读边界
- 所据为版本未注明的索引摘要，文献是讨论稿而非主论文；不能把提问当作已证实缺陷，也不能据此确认机器学习或贝叶斯方法更优。
原文摘要
According to Leo Breiman (2001), there are two broad cultures for analysing and modelling to reach conclusions from the data. The first one is data modelling culture where the value of the parameters are estimated from the data and then the model is used for information and/or prediction. The second one is algorithmic modelling culture, where the approach is to find a function f(x) using an algorithm that operates on x to predict y. In the first case, statistical tools like OLS are used whereas machine learning tools perform much better in the second case. I should congratulate the authors for developing a hybrid model using machine learning and traditional statistical techniques like OLS. However, I do have few queries on the real-life applications of the proposed model. Most of the simulation results for main effects, effect modification and misspecified link function are valid for sample size n = 500 or more. Also, the data analysis example presented in section 7 has a sample size of 2500. How is the empirical bias affected when the sample size is small? Bayesian models incorporate prior information, which can be specified quantitatively in the form of a distribution. Do the authors think a Bayesian parameter generalised linear model can perform better if some prior information is available? Also, I am not sure how this model tackles the problem of multicollinearity? Finally, in the context of high dimensional variable selection, what will happen if the covariates are not distributed normally?