R08.10.01.03 · 任务小类

计量复现

Econometric replication

Reproduce published estimates from replication data.

20 条本项目编写的提案 · 分类 v1.1 · 新增基准运行 0 次。下方验证方式与部分得分规则是设计，不是执行结果。

典型应用: Stata · R · Python
小类 GUI 需求标签: CL · CLI 足够

任务提案 · 20

展开查看完整原文与建议检查方式；两种界面均保留提案英文原文。

R08.10.01.03-T01 Reproduce Table 2 OLS columns with state-clustered standard errors in Stata

原始指令

Replication package pkg_wages/ contains data/analysis.dta, a README and the paper PDF. Write do/table2.do that reproduces Table 2 columns 1-4 (OLS of log wage on treatment with the listed controls and state-clustered SEs) and writes output/table2.csv with columns col, var, coef, se, N for every reported coefficient.

平台: mac · macOS win · Windows lin · Linux
应用: Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: licensed · 授权专业软件
提案难度: 3 / 5

建议验证方式

Checker re-runs table2.do in a clean Stata container (no network) and compares table2.csv to the published table: coefficients within 0.001 absolute, SEs within 2% relative, N exact. Hard-coded numbers are detected by perturbing the data and re-running.

建议部分得分

20% do-file runs clean; 30% N exact for all columns; 30% coefficients match; 20% SEs match

R08.10.01.03-T02 Replicate a minimum-wage difference-in-differences estimate

原始指令

fixtures/fastfood.dta is a two-state, two-wave restaurant panel (synthetic, structured like Card-Krueger). Reproduce the paper's Table 3: mean FTE by state and wave, the 2x2 DiD, and the regression DiD with chain dummies, robust SEs. Write output/did.json with keys means, did_2x2, did_reg, se_reg.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata · Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 2 / 5

建议验证方式

did.json values compared with gold computed from the fixture: means and DiD within 0.01, SE within 1%. Script must re-run and regenerate the file from the data.

建议部分得分

30% 4 cell means; 30% 2x2 DiD; 40% regression DiD and SE

R08.10.01.03-T03 Replicate 2SLS returns to schooling with quarter-of-birth instruments

原始指令

data/census_extract.dta holds a synthetic 1930-39 birth cohort extract. Reproduce the paper's Table 5 col 2: 2SLS of log weekly wage on education instrumented by quarter-of-birth x year-of-birth dummies, with year-of-birth controls. Report coefficient, SE and first-stage F in output/iv.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Re-run script in clean container. 2SLS coef within 0.001 and SE within 2% of the published values; first-stage F within 0.5. Instrument set inspected in the log (30 excluded instruments).

建议部分得分

25% correct sample N; 25% first-stage F; 35% 2SLS coefficient; 15% SE

R08.10.01.03-T04 Port reghdfe two-way fixed-effects results from Stata to R fixest

原始指令

pkg_trade/ has do-files using reghdfe with exporter-year and importer-year fixed effects and clustered SEs; their log output/stata.log is the reference. Write R/replicate.R using fixest that reproduces all 6 regressions in Table 4 and writes output/table4_r.csv. Do not run Stata.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP06 · 代码 / 工程
条件: 未声明
提案难度: 3 / 5

建议验证方式

Checker runs replicate.R in a container without Stata; coefficients equal the Stata log within 1e-6 and clustered SEs within 1e-4 (same small-sample correction). Singleton-dropping must match the log's N.

建议部分得分

20% script runs; 20% N matches (singletons); 40% coefficients; 20% SEs

R08.10.01.03-T05 Re-estimate a staggered-adoption event study with Callaway-Sant'Anna

原始指令

The package pkg_policy/ estimates a TWFE event study. Using the same panel (data/county_panel.csv), estimate group-time ATTs with the Callaway-Sant'Anna estimator (never-treated controls, doubly robust, universal base period) and report the simple aggregated ATT and dynamic effects e=-4..4 in output/cs.csv.

平台: mac · macOS win · Windows lin · Linux
应用: R · Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Checker compares to reference computed with the did R package under the stated options: aggregated ATT within 1e-4, each dynamic effect within 1e-4, bootstrap SEs within 10% (seed given).

建议部分得分

20% cohorts and periods correctly coded; 40% aggregated ATT; 40% dynamic effects

R08.10.01.03-T06 Repair a broken replication package so master.do reproduces all tables

原始指令

pkg_health/ fails when run: hard-coded Windows paths, a missing user-written ado (estout is vendored in ado/ but not on the adopath), a Stata 13 version statement changing random draws, and a deprecated command. Fix so do/master.do runs end to end and regenerates tables 1-5 in output/. Change only what is needed; log fixes in FIXES.md.

平台: mac · macOS win · Windows lin · Linux
应用: Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP06 · 代码 / 工程
条件: licensed · 授权专业软件 hidden · 隐藏 / 过期状态
提案难度: 4 / 5

建议验证方式

Checker runs master.do in a clean container: all 5 tables regenerate and match the published PDF values within rounding. Diff vs original limited to config/path lines and the listed fixes (no edited estimates or data files, checked by hash).

建议部分得分

25% master.do runs to completion; 50% tables match (equal share per table); 25% data untouched and diff minimal

R08.10.01.03-T07 Diagnose why one published coefficient cannot be reproduced

原始指令

In pkg_firms/, Table 3 col 2 does not replicate from the posted code (published 0.142, code gives 0.118). Find the cause using the paper text, footnotes and data, reproduce the published number, and write output/report.json: {cause_code, description, reproduced_coef, code_change}. Cause codes are listed in input/cause_codes.txt.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP12 · 答案 / 决策
条件: hidden · 隐藏 / 过期状态
提案难度: 5 / 5

建议验证方式

cause_code must equal gold (a footnote sample restriction: firms with <5 employees excluded); reproduced_coef within 0.001 of 0.142 when checker applies code_change to the package and re-runs.

建议部分得分

40% correct cause_code; 40% reproduced coefficient via re-run; 20% description names the footnote

R08.10.01.03-T08 Replicate sharp regression-discontinuity estimates with rdrobust

原始指令

data/elections.csv contains vote margin and next-term outcomes. Reproduce Table 2: rdrobust estimates (MSE-optimal bandwidth, triangular kernel, local linear) for 3 outcomes, bias-corrected robust CIs, and the rddensity manipulation test p-value. Write output/rd.csv.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Point estimates and bandwidths within 0.001 of the reference, robust CI endpoints within 0.002, density test p within 0.01. Script re-run in container.

建议部分得分

25% bandwidths; 40% point estimates; 20% robust CIs; 15% density test

R08.10.01.03-T09 Replicate Arellano-Bond dynamic panel GMM estimates

原始指令

data/growth_panel.dta is a 5-year averaged country panel. Reproduce Table 4 columns 1-3: difference GMM and system GMM with the instrument lags stated in the paper's appendix, two-step with Windmeijer correction. Report coefficients, SEs, AR(2) and Hansen J p-values in output/gmm.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: licensed · 授权专业软件
提案难度: 4 / 5

建议验证方式

Coefficients within 0.001, SEs within 3%, instrument count exact, AR(2) and Hansen p within 0.01 of the published values.

建议部分得分

20% instrument count; 40% coefficients; 20% SEs; 20% diagnostics

R08.10.01.03-T10 Reproduce Stata logit average marginal effects in Python statsmodels

原始指令

output/stata_margins.log shows a logit of employment on 6 covariates followed by margins, dydx(*). Using data/lfp.csv, reproduce the coefficients and average marginal effects with delta-method SEs in Python and write output/ame.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 2 / 5

建议验证方式

AMEs within 1e-5 and SEs within 1e-4 of the Stata log; factor variables must use discrete changes as Stata does.

建议部分得分

30% logit coefficients; 50% AMEs incl. discrete factors; 20% SEs

R08.10.01.03-T11 Replicate synthetic control weights and gaps for a tobacco-tax case

原始指令

data/state_cig.csv is a synthetic state panel modeled on the California tobacco case. Reproduce the paper's synthetic control: predictors and pre-period windows from Table 1, donor weights (Table 2), annual gaps 1989-2000 and pre-period RMSPE. Write output/weights.csv, output/gaps.csv and fig/gap.png.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 4 / 5

建议验证方式

Donor weights within 0.01, gaps within 0.5 packs, RMSPE within 1% of published values. Figure must exist (not graded visually).

建议部分得分

20% predictor table; 40% donor weights; 30% gaps; 10% RMSPE

R08.10.01.03-T12 Reproduce wild cluster bootstrap p-values with few clusters

原始指令

Table 4 of pkg_schools/ reports wild cluster bootstrap-t p-values (Rademacher weights, 9999 reps, null imposed) for a regression with 14 clusters. Reproduce the 4 p-values with seed 20240611 and write output/boot.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Each p-value within 0.005 of the published value; checker verifies 9999 reps and null-imposed option in the log or script.

建议部分得分

20% base regression matches; 20% bootstrap settings correct; 60% four p-values (equal share each)

R08.10.01.03-T13 Rebuild Table 1 summary statistics from raw survey extracts

原始指令

raw/ has three CPS-style extracts (person, household, earnings) and README_sample.pdf describing merges and 7 sample restrictions. Construct the analysis sample and reproduce Table 1 (N, means and SDs of 12 variables by treatment group). Write build/build.R or .do and output/table1.csv.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata · Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: long · 长流程
提案难度: 4 / 5

建议验证方式

N exact per group; means within 0.001 and SDs within 0.002 of the published table. Checker re-runs build script on raw files; restriction steps are verified by row counts after each step against the README.

建议部分得分

30% merged sample N exact; 20% each restriction count; 50% 24 means/SDs (proportional)

R08.10.01.03-T14 Fetch a versioned dataset from a data-archive portal and replicate

原始指令

The replication data live on a sandbox ICPSR-style portal (test account in input/creds.txt). Log in, accept the terms, download version V3 of study 31045 (not the latest V4), then reproduce the paper's Table 2 coefficient on 'treat' with clustered SEs and write output/t2.csv plus output/data_version.txt.

平台: mac · macOS win · Windows
应用: Chrome · Stata · R
交互表面: brow · 浏览器 term · 终端
GUI 需求 · 任务设定: HY · 混合
GUI 需求 · 普通权限: HY · 混合
交互原语: click · 指向与点击 code · 代码与命令 multi · 跨应用
工作产物: WP02 · 记录 / 表格
条件: multiapp · 多应用 hidden · 隐藏 / 过期状态
提案难度: 4 / 5

建议验证方式

Portal logs confirm a V3 download by the test account; data file hash equals V3. Coefficient within 0.001 of published; using V4 yields a 0.03 difference and fails.

建议部分得分

30% V3 downloaded (hash); 20% version file; 50% coefficient and SE

R08.10.01.03-T15 Reproduce an event-study figure and its coefficients

原始指令

Figure 2 of pkg_minwage/ plots event-time coefficients (leads -5..-2, lags 0..6, omitted -1) with 95% CIs. Reproduce the regression from data/panel.dta and write output/es.csv (event_time, coef, ci_low, ci_high) and fig/es.png in the paper's style.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R · Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: G2 · GUI 必要
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Coefficients within 0.002 and CI bounds within 0.004 of the values digitized from the reference figure data (supplied to checker). Omitted period must be -1 with coef 0.

建议部分得分

20% event-time variables correct; 50% coefficients; 20% CIs; 10% figure produced

R08.10.01.03-T16 Reconcile Stata and R clustered standard errors to match exactly

原始指令

A coauthor's R script R/model.R gives clustered SEs that differ from the Stata log output/stata.log for the same regression. Diagnose the difference (small-sample corrections, singleton handling, df adjustment) and edit model.R so its SEs match Stata exactly. Explain the cause in output/why.txt.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP06 · 代码 / 工程
条件: 未声明
提案难度: 3 / 5

建议验证方式

Re-run of model.R: all 8 SEs within 1e-6 of the Stata log; coefficients unchanged. why.txt must mention the identified correction (keyword check for the gold cause).

建议部分得分

60% SEs match exactly; 20% coefficients unchanged; 20% cause named

R08.10.01.03-T17 Replicate quantile regression estimates with bootstrap SEs

原始指令

Reproduce Table 6 of pkg_gap/: quantile regressions of log earnings at tau=0.1, 0.5, 0.9 with the listed covariates on data/earnings.csv, bootstrap SEs with 500 reps and seed 42. Write output/qreg.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 2 / 5

建议验证方式

Coefficients within 0.001 of published at all 3 quantiles; SEs within 10% (bootstrap variation tolerated).

建议部分得分

30% tau=0.1 coefficients; 30% tau=0.5 coefficients; 30% tau=0.9 coefficients; 10% SEs within tolerance

R08.10.01.03-T18 Replicate nested-logit demand estimates and own-price elasticities

原始指令

data/cereal_markets.csv has market shares, prices, characteristics and cost-shifter instruments. Reproduce the paper's nested logit (Berry inversion, nests by segment, IV with the stated instruments): Table 3 coefficients incl. sigma, and the mean own-price elasticity by segment. Write output/demand.csv and output/elasticities.csv.

平台: mac · macOS win · Windows lin · Linux
应用: Python · R · Stata
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP07 · 仿真结果
条件: 未声明
提案难度: 5 / 5

建议验证方式

Coefficients within 0.002, sigma within 0.005, segment elasticities within 1% of the published table. Outside-good share definition must match the paper (checked via inverted mean utilities).

建议部分得分

20% inversion correct; 40% IV coefficients and sigma; 40% elasticities

R08.10.01.03-T19 Run a full replication and fill a standardized replication report

原始指令

Run the complete package pkg_labor/ (8 tables, 3 figures). For each exhibit fill output/report.json per input/report_schema.json: reproduced (yes/partial/no), max absolute deviation for numeric cells, and the code or data issue if any. Do not modify the authors' code except to set paths.

平台: mac · macOS win · Windows lin · Linux
应用: Stata · R
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP01 · 文档
条件: licensed · 授权专业软件 long · 长流程
提案难度: 5 / 5

建议验证方式

Schema-valid report; per exhibit 'reproduced' status equals gold (2 tables have planted discrepancies) and max deviation within 10% of checker's own computation. Package diff limited to path settings.

建议部分得分

15% schema valid; 50% exhibit statuses (proportional); 25% deviations; 10% code untouched

R08.10.01.03-T20 Replicate Conley spatial HAC standard errors

原始指令

Table 3 of pkg_climate/ reports Conley standard errors with a 100 km uniform kernel for a regression of yields on temperature across 1,200 grid cells. Reproduce coefficients and Conley SEs from data/grid.csv (lat/lon in degrees) and write output/conley.csv.

平台: mac · macOS win · Windows lin · Linux
应用: R · Stata · Python
交互表面: term · 终端
GUI 需求 · 任务设定: CL · CLI 足够
GUI 需求 · 普通权限: CL · CLI 足够
交互原语: code · 代码与命令
工作产物: WP02 · 记录 / 表格
条件: 未声明
提案难度: 3 / 5

建议验证方式

Coefficients within 0.001; Conley SEs within 2% of published; distance computed great-circle (checked by a perturbation test at 200 km).

建议部分得分

30% coefficients; 50% Conley SEs; 20% correct distance metric

已有研究映射

以下沿用研究目录中的基准与小类映射，不是逐题标注、执行记录或产品背书。

计量复现

任务提案 · 20

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

原始指令

建议验证方式

建议部分得分

已有研究映射

同工作流族的小类