Dissertation replication and workflow guide

Materials overview

Dissertation material Reader-facing destination Main source or saved output
Page 2: flood-risk coverage Background calculations scripts/pull_distinct_resPINs.R, scripts/pull_floodfactor_pin_list.R
Page 35 / Appendix A Q1 assumptions Q1 page, historical q1_assumption_tests.qmd, scripts/Q1_SFHA_DiD_test.R
Page 56 / Tables XIII-XIV Exposure comparisons Q1_Model_Comparison.qmd, etwfe_combined_mega_grid.qmd, saved grid tables
Page 62 Q1 robustness Nearby-controls Quarto page and prepared building data
Page 69 / Appendix B Q2 diagnostics scripts/q2_assumption_tests/, scripts/Q2_FloodFactor_didFF_test.R
Page 77 / Appendix C Q2 coefficients scripts/q2_dissertation_scripts/, outputs/q2_marginaleffects/

Additional analyses include Flood Factor variable-form checks and the file inventory. The inventory contains historical entries as well as active files; use actual paths and the limitations below when reproducing results.

Repository organization

Location Role
_targets.R, R/ Spatial and sales-data construction and helper functions
inputs/, data/raw/ Spatial inputs, sales exports, lookups, and acquisition inputs
data/processed/targets/sales/ Current allsales_* exports alongside older df_prep_* analysis inputs
data/data_used_in_dissertation/ Saved data snapshots; its README identifies March 8, 2026 sales data
scripts/q2_dissertation_scripts/ Numbered preparation, model, marginal-effect, and robustness scripts
scripts/q2_FF_variable_checks/ Binary, ordinal, and continuous score comparisons
scripts/q2_assumption_tests/ Saved diagnostic plot/model construction
outputs/ Model objects, tables, figures, maps, and diagnostics
Root .qmd files Main website pages and additional exploratory analyses
_quarto.yml, docs/ Website build configuration and generated HTML
_targets/, _freeze/, .quarto/ Target objects and execution/build caches; not source data
archived/, quarto_supportfiles/ Historical analyses and rendered materials

Data and software

Use the dissertation project root as the working directory and begin with a fresh R session. The inspected project has no renv.lock, so the original package environment is not pinned. R and Quarto are required. Data processing uses packages including targets, tarchetypes, sf, tidyverse, readxl, DBI, and RSQLite; analyses and reports additionally use fixest, did, etwfe, marginaleffects, didFF, broom, modelsummary, gt, flextable, DT, and knitr. Check each source file’s imports and record sessionInfo() for a reproducibility run.

The inputs combine Cook County sales and parcels, FEMA flood boundaries and map timing, Microsoft building footprints, and First Street Flood Factor information. The current targets configuration selects data/raw/Assessor_-_Parcel_Sales_20260308.csv. The residential-PIN acquisition script uses a separately located PTAXSIM database; targets also reads a municipality lookup outside this repository. Manual GIS buffers, corrections, and parcel exclusions are additional inputs. Exact acquisition dates, permitted redistribution, and a complete input manifest remain to be documented; local availability is not the same as public availability. See the background source notes.

Computational workflow

External source data + manual GIS/lookup inputs
  -> _targets.R + R/ helpers
  -> processed spatial and sales exports
  -> analysis preparation / Q1 Quarto models / Q2 numbered scripts
  -> saved model, contrast, table, and diagnostic objects
  -> selected Quarto pages
  -> docs/ HTML
  -> separate publication step
  1. Inventory inputs and versions. Preserve the dissertation snapshots and identify which saved output belongs to which source vintage. Do not replace a df_prep_* input with allsales_* solely because its name is newer.
  2. Inspect the data pipeline. From the root, targets::tar_manifest() lists declared targets and targets::tar_outdated() identifies recorded changes. A later targets::tar_make() rebuilds affected targets after input paths are resolved. Some files are read inside commands rather than declared as file targets; these require explicit attention when inputs change.
  3. Prepare analysis inputs. The current pipeline exports allsales_*; downstream models still require saved df_prep_* inputs. Restoring the reproducible bridge between these is a prerequisite for a clean build.
  4. Generate analysis results. Q1 models are distributed across Quarto files. Q1_Model_Comparison.qmd writes the TWFE grid; etwfe_combined_mega_grid.qmd assembles the ETWFE grid from saved results. For Q2, the intended order in scripts/q2_dissertation_scripts/ is 00 preparation, 01 basic TWFE, 02 ETWFE, 03 main marginal effects, and 04 robustness. The Flood Factor form-check sequence is 00, 01, 02, with 03 a loading example. The numbered script source paths have been repaired; verify the saved model version and external inputs before running these sequences.
  5. Prepare diagnostics. Existing Q2 plot and model RDS files are under outputs/q2_assumption_tests/. Their runner now reuses the existing prepared dataset or calls the existing Flood Factor preparation script when the dataset is absent. The website can display saved objects, but a clean diagnostic rebuild is not yet documented end to end. Historical distributional-test scripts contain exploratory alternatives; do not run them indiscriminately, because some overwrite outputs.
  6. Render and inspect. The website explicitly renders selected pages to docs/. For a chosen page, use quarto render filename.qmd; use quarto render for the configured site. Some pages fit models during rendering. The freeze: auto setting can reuse cached execution in project builds, so external RDS changes must be considered when refreshing output. Inspect tables, labels, links, and figure provenance before publication.
  7. Publish separately. Confirm the approved local sources and outputs are versioned and publish through the project’s chosen GitHub Pages workflow. Preserve the stable homepage URL used by the dissertation. This documentation edit does not publish the site.