First Course Ding

This page is the working table of contents for a crabbymetrics translation pass over the Peng Ding notebooks in ding_w_source.

1 Current Batch

The first reviewable batch is already underway:

Each grouped section links out to the chapter-level pages already living under docs/ding/.

The working rules for the port are:

  • use crabbymetrics estimators and primitives whenever the chapter logic allows it
  • keep external dependencies minimal: numpy, matplotlib, and pandas or polars only when a CSV or Stata read is genuinely required
  • avoid statsmodels, sklearn, scipy, linearmodels, and similar notebook-time dependencies in the translated docs unless a chapter is blocked on a missing crabbymetrics feature
  • prefer one Quarto page per chapter, with a small number of section pages to group completed chapters in the navbar

2 Implementation Batches

The rough order is:

  1. Randomized-experiment foundations and design-based inference.
  2. Observational studies and semiparametric estimators.
  3. IV and fuzzy-RD chapters.
  4. Principal stratification, mediation, and any residual appendix material.

This ordering matches the current library surface. The earliest chapters mostly need numpy, plotting, and some OLS or randomization-inference utilities. The middle chapters map onto BalancingWeights, AIPW, PartiallyLinearDML, EPLM, and AverageDerivative. The later IV chapters fit naturally on top of TwoSLS and GMM. The biggest likely blockers are matching, local-polynomial RD, principal stratification, and mediation.

3 Planned TOC

Chapter Source notebook Planned docs page crabbymetrics spine Minimal deps Notes
1 Chapter01CorrAssocSimpsons.ipynb ding/ch01-correlation-simpson.qmd summaries + OLS where useful numpy, pandas, matplotlib implemented
2 Chapter02PotentialOutcomes.ipynb ding/ch02-potential-outcomes.qmd numpy estimand algebra and simulation numpy, matplotlib implemented
3 Chapter03CREandFRT.ipynb ding/ch03-cre-frt.qmd difference-in-means, OLS, permutation/randomization logic numpy, matplotlib implemented
4 Chapter04CREandNeyman.ipynb ding/ch04-cre-neyman.qmd design-based variance calculations + simulation numpy, matplotlib implemented
5 Chapter05StratandPostStrat.ipynb ding/ch05-stratification.qmd weighted means, post-stratification, blocked OLS numpy, pandas, matplotlib implemented
6 Chapter06RegadjRerand.ipynb ding/ch06-regadj-rerand.qmd centered OLS, Lin-style adjustment, rerandomization simulation numpy, pandas, matplotlib implemented
7 Chapter07MatchedPairs.ipynb ding/ch07-matched-pairs.qmd paired means and exact sign-flip randomization numpy, pandas, matplotlib implemented
8 Chapter08UnifyingFisherNeyman.ipynb ding/ch08-fisher-neyman.qmd randomization and repeated-sampling simulations numpy, matplotlib implemented
9 Chapter09BridgingFinitePopAndSuperPop.ipynb ablations/bridging-finite-and-superpopulation.qmd OLS + stacked GMM numpy, matplotlib already implemented
10 Chapter10ObsStudiesSelBias.ipynb ding/ch10-selection-bias.qmd observational-study simulation + balance diagnostics numpy, matplotlib natural next chapter after the current observational batch
11 Chapter11Pscore.ipynb ding/ch11-propensity-score.qmd Logit, BalancingWeights numpy, pandas, matplotlib implemented
12 Chapter12DoubleRobustATE.ipynb ding/ch12-double-robust-ate.qmd AIPW, Logit, OLS numpy, pandas, matplotlib implemented
13 Chapter13DoubleRobustATT.ipynb ding/ch13-double-robust-att.qmd BalancingWeights, ATT weighting, outcome adjustment numpy, pandas, matplotlib implemented
14 none in source none none none no chapter file present
15 Chapter15Matching.ipynb ding/ch15-matching.qmd nearest-neighbor matching numpy, pandas, matplotlib likely blocked on a small matching helper in crabbymetrics
16 Chapter16UnconfDifficulties.ipynb ding/ch16-unconfoundedness.qmd overlap and model-misspecification simulations numpy, pandas, matplotlib should be possible without new estimators
17 Chapter17Evalue.ipynb ding/ch17-evalue.qmd analytic sensitivity summaries numpy, matplotlib probably wants a dedicated helper but simple enough
18 Chapter18SensitivityAnalysis.ipynb ding/ch18-sensitivity-analysis.qmd omitted-confounding sensitivity calculations numpy, pandas, matplotlib may justify a reusable sensitivity module
19 Chapter19RosenbaumPvalues.ipynb ding/ch19-rosenbaum.qmd matched-study sensitivity and p-values numpy, pandas, matplotlib blocked on matching-set support and Rosenbaum-style routines
20 Chapter20OverlapRD.ipynb ding/ch20-overlap-rd.qmd overlap diagnostics and RD plots numpy, pandas, matplotlib a full local-polynomial RD estimator is likely a new feature
21 Chapter21IVexperiments.ipynb ding/ch21-iv-experiments.qmd Wald estimands, TwoSLS, compliance simulations numpy, matplotlib implemented
22 Chapter22IVmixtureDist.ipynb ding/ch22-iv-inequalities.qmd IV bounds and mixture-distribution logic numpy, matplotlib mostly array algebra and plotting
23 Chapter23IVeconometrics.ipynb ding/ch23-iv-econometrics.qmd TwoSLS, GMM numpy, pandas, matplotlib implemented
24 Chapter24IVfuzzyRD.ipynb ding/ch24-fuzzy-rd.qmd fuzzy RD as IV numpy, pandas, matplotlib likely blocked on a local-polynomial RD helper
25 Chapter25IVmendelian.ipynb ding/ch25-mendelian-randomization.qmd ratio and multi-instrument TwoSLS numpy, pandas, matplotlib feasible with current IV machinery
26 Chapter26principalStratification.ipynb ding/ch26-principal-stratification.qmd latent-strata models numpy, pandas, matplotlib probably needs new library support before a real port
27 Chapter27mediationAnalysis.ipynb ding/ch27-mediation.qmd mediation via sequential regressions / g-computation numpy, pandas, matplotlib likely needs new helpers or a scoped estimator
A ChapterA.ipynb optional ding/appendix.qmd formulas and helper notes numpy low priority unless later chapters depend on it

4 Suggested Next Steps

The next concrete implementation batch should be:

  1. Chapter 10 to bridge the randomized and observational sections.
  2. Chapters 15 through 20 once matching and sensitivity helpers are scoped.
  3. Chapters 22, 24, and 25 as the remaining IV material.
  4. Chapters 26 and 27 only after the necessary latent-structure and mediation support exists.