把一篇笔记能用到的部件全部摆在同一页上 —— 页头出自 frontmatter、导语、三种框、表格与判定标签、代码块、插图、页脚收口 —— 每个都标出它在页面的位置和归属的色相。写新笔记时对着这一页抄即可。
Site notes
2Written directly in this repository目录约定、frontmatter 字段、版式从哪来、本地预览与发布，以及那 9 个独立项目站现在放在哪里。
SAS patterns
2De-identified, reusable code shapesResolving a replacement chain of unknown depth in one pass of a DATA step: why an in-memory hash is the only structure that fits this shape of problem, what the step guarantees, and what it does when the extract is not clean. A self-contained Base SAS check ships with the note — one fixture, eleven assertions.
一次把整张表读进内存，再让主表的每一行各自去“钉”上去——这是临床 SAS 编程里最值钱的一条肌肉记忆。 本文给出 hash 对象的完整语法骨架（declare → definekey → definedata → definedone → find / output），四类真实项目形态的 可搬运代码（受试者筛选、横向合并、纵向检索、SDTM/ADaM 衍生），以及八条按官方文档与会议论文归集的易错点。
Languages and practice
1A multi-language path out of SAS把前端拆成「结构 — 表现 — 行为 — 运行环境 — 数据通路 — 工程交付」六层，逐层给出核心概念、术语、 最小示例与验收标准，并配知识结构图与流程图。写作语境按读者现有的 R / Python / SAS 心智模型来搭桥， 终点是能独立做出一个可用的页面并放到网上。
Learning log
1Built in the openLife and travel
1Plans and references outside workAll notes
7Newest firstExternal projects
11Each keeps its own siteA learning path from matrix foundations to regulatory submission, covering covariance structures, REML and Kenward-Roger, implementations in SAS and R, and ICH E9(R1).
West, Welch and Galecki, chapter by chapter, read for a clinical statistical programmer: what each concept corresponds to in a trial, the SAS that fits it, and where it bites. Six case studies mapped onto multicentre trials, repeated visits, crossed raters and design effect.
The production pattern behind a Phase III primary endpoint: PROC MIXED with no RANDOM statement, TYPE=UN and DDFM=KR, the baseline record leaving the response vector, what each covariance structure costs, and the prespecification in the SAP the model cannot supply for itself. Ships a runnable Base SAS check.
How eighteen shipped system prompts are actually written, measured across the seven dimensions they all answer — and the argument the note turns on: every 'never' should be answerable with which layer actually stops the model when it disobeys. Written in Chinese, with twenty-three quoted excerpts, each checked verbatim against the archive it came from.
Cumulative incidence against Kaplan-Meier: transformation-based and inverted confidence intervals, the PROC LIFETEST option family taken apart, and runnable SAS.
De-identified write-ups of coding patterns from clinical statistical programming, each a self-contained HTML file with no dependencies. Includes the SAS hash object note, in Chinese.
For programmers fluent in SAS and new to Python: chapters and runnable case studies on public CDISC data, from syntax to building and deploying AI agents.
An R course for clinical SAS programmers: CDISC data, statistical analysis, tables and figures, package development and Shiny, with worked studies on public data.
One search across clinical programming conference papers from PharmaSUG, PHUSE, SAS Global Forum, SAS Innovate, R/Pharma and others.
Trending, topic and rising-star rankings rebuilt each week from scraped trend data, with the underlying data archived over time.
An open learning log on agent development: write a minimal framework first to establish the underlying principles, dissect production open-source projects next, then apply what holds up to clinical work.