Household survey analytics
Defines household and person units separately, reports denominators, builds housing and vulnerability indicators, and documents limits of interpretation.
Privacy-first household survey analytics
Rural Bolivia Housing Analytics
Privacy-first household survey analytics for understanding rural housing conditions, service access and territorial inequalities in Bolivia.
A reproducible public workflow using synthetic household and person records to demonstrate Python data processing, SQL modelling, validation, housing indicators, uncertainty reporting and policy-oriented communication without exposing respondent records.
Public data are syntheticOriginal identifiable records are not distributed in the public repository.
Data analyst stackPython pipeline, SQL analytical model, tests, reports, figures and GitHub Pages publication.
Interpretation disciplineAssociational estimates, uncertainty intervals and missingness are reported before substantive claims.
Executive summary
What the project demonstrates
This repository is a public, privacy-preserving version of a household-survey analysis workflow. It separates private research records from public communication by using fully synthetic data to demonstrate validation, uncertainty reporting, modelling and publication.
Privacy-first publication
Uses synthetic public records, excludes original identifiable files, avoids publishing direct identifiers, and documents responsible use.
Recruiter-facing evidence
Shows Python, SQL, validation tests, analytical reporting, visual communication and reproducibility with a fixed-seed workflow.
Key metrics
Verified public workflow figures
All figures below are already documented in the repository outputs. They describe the synthetic public workflow and must not be interpreted as estimates for Coroico or Bolivia.
60Synthetic households
271Synthetic household members
37.9%Weighted adequate-housing estimate
26.7%-51.7%Bootstrap 95% interval
BOB 2210Median monthly household income
15.0%Income missingness
44.6Mean vulnerability index
10Publication figures available
Key findings
Documented insights from the existing outputs
The findings are methodological and illustrative because the public records are synthetic. No causal claims are made.
Housing adequacy is reported with uncertainty
The synthetic workflow reports a weighted adequate-housing estimate of 37.9% with a 95% bootstrap interval of 26.7%-51.7%.
Service access is part of the housing score
The housing score combines electricity, improved water, improved sanitation, wall quality, floor quality and severe-crowding status.
Missingness is visible before modelling
Income missingness is reported as 15.0%, and the missing-data audit is published as a first-class analytical output.
Territorial comparisons are cautious
Zone-level housing outputs include intervals, making the uncertainty around small samples visible instead of encouraging unstable rankings.
Regression remains exploratory
The logistic model is parsimonious for a 60-household sample, and the documented model intervals cross the null.
Privacy shapes the publication layer
Public outputs are synthetic, aggregate or methodological, with original identifiable records excluded from the repository.
Dashboard or report
Interactive analytical report
The repository contains a public GitHub Pages report and a STROBE-aligned HTML paper. No separate dashboard was rebuilt in this phase.
Public analytical report
Web-based publication layer presenting the synthetic workflow, headline metrics, figures, methods, privacy statement and links to reproducible assets.
Main sections
- Executive summary and verified metrics.
- Housing adequacy, service access, missingness and vulnerability figures.
- Executive tables from the HTML paper and existing CSV outputs.
- Workflow, methodology, privacy, reproducibility and limitations.
Main figures
Existing visual evidence
Eight non-redundant figures are reused from `docs/figures`. They summarize data quality, income, housing, service access, gender outcomes, vulnerability and the exploratory model.
Missing data audit
Period: synthetic cross-sectional workflow. Unit: variable-level missingness. Source: synthetic public data. Method note: missingness is disclosed before outcome modelling.
Income distribution
Period: synthetic cross-sectional workflow. Unit: households. Source: synthetic public data. Method note: income is summarized for descriptive context only.
Housing adequacy by synthetic zone
Period: synthetic cross-sectional workflow. Unit: households. Source: synthetic public data. Method note: intervals are shown to avoid over-ranking small groups.
Service access
Period: synthetic cross-sectional workflow. Unit: households. Source: synthetic public data. Method note: service access enters the transparent housing measurement logic.
Gender-disaggregated outcomes
Period: synthetic cross-sectional workflow. Unit: persons and households as documented. Source: synthetic public data. Method note: differences are descriptive.
Income and vulnerability
Period: synthetic cross-sectional workflow. Unit: households. Source: synthetic public data. Method note: vulnerability is analytically defined and illustrative.
Correlation heatmap
Period: synthetic cross-sectional workflow. Unit: variables. Source: synthetic public data. Method note: associations are exploratory and not causal.
Exploratory model coefficients
Period: synthetic cross-sectional workflow. Unit: household model. Source: synthetic public data. Method note: odds ratios summarize association, not causal effects.
Executive tables
Reusable tabular outputs
Existing tables are linked without regeneration. Files outside `/docs` are linked through GitHub so they remain reachable when Pages is served from `/docs`.
Table 1. Synthetic sample characteristics
Households, members, income, missingness, adequate housing and vulnerability summary.
Open tableTable 2. Exploratory logistic model
Odds ratios and intervals for education, log income, household size and woman-headed household.
Open tableModel results CSV
Machine-readable model coefficients already stored in the reports directory.
Open CSVSTROBE reporting map
Checklist-style table mapping reporting items to the public methodological paper.
Open checklistData sources
Privacy-preserving public data layer
The public repository contains synthetic, non-identifiable records and documentation. Original identifiable records are excluded from public distribution.
Household survey microdata
Represented publicly through synthetic household and member files generated for reproducibility and privacy.
Public products
Figures, HTML report, CSV model results, source code, tests and metadata are published for review.
Restricted material
Original identifiable microdata are not distributed, and public outputs must not be used as population estimates.
Analytical workflow
From private records to public outputs
Household survey microdata
Private source records are kept outside the public publication layer.
Privacy and validation checks
Data contracts reject direct identifiers and invalid ranges before analysis.
Household and housing-variable harmonization
Household and member units are documented separately to protect denominators.
Rural housing indicators
Service access, quality materials and severe crowding feed the housing score.
Territorial comparisons
Zone comparisons include uncertainty to prevent unstable rankings.
Figures, tables, reports and dashboard
Outputs are communicated through GitHub Pages, the HTML paper, figures and reproducible assets.
Methodology
Existing analytical design
The methods below summarize the repository's existing documentation and report outputs. Nothing was recalculated for this web update.
Variable construction
Housing score combines electricity, improved water, improved sanitation, quality walls, quality floor and no severe crowding.
Weights and uncertainty
Weighted descriptive estimates and 2,000-resample bootstrap intervals are documented in the technical report.
Missing values
Missingness is audited before modelling, with income missingness reported in the public metrics.
Territorial analysis
Zone-level comparisons are used for communication, with wide intervals treated as a substantive warning.
Exploratory model
A logistic model relates adequate housing to education, log income, household size and woman-headed household.
Limits
Synthetic records, small sample size and cross-sectional associations restrict substantive and causal interpretation.
Privacy
Responsible public release
The repository is designed to demonstrate analytical skill while avoiding disclosure of real respondents.
No public microdata disclosure
The original identifiable household records are not distributed publicly.
Public products are safe
Published files are synthetic, aggregate, methodological or code assets.
No identifiers
The test suite checks that public generated columns avoid forbidden direct identifiers.
Use restrictions documented
Public outputs are not valid estimates for Coroico, Bolivia or any real population.
Privacy controls visible
Validation, fixed-seed generation and responsible-use documentation are part of the repository.
Manual review still matters
A real survey release would require sampling design, field documentation and authorization before substantive publication.
Reports
Existing research communication products
STROBE-aligned research paper
HTML paper with abstract, methods, results, figures, two embedded tables, limitations and declarations.
Open paperExecutive summary
Short decision-oriented summary of the synthetic workflow, headline metrics and limitations.
Open summaryTechnical report source
Quarto source documenting the research design, estimands, measurement, missingness, descriptive outputs and model specification.
Open sourceReproducibility
Code, tests and analytical assets
The repository documents commands and contains Python source, tests, SQL and outputs. This web update did not run pipelines or regenerate results.
Python workflow
Source files generate synthetic records, validate data contracts, produce metrics, fit the exploratory model and publish outputs.
SQL analytical model
The SQL file builds a portable household-member analytical view with quality flags and synthetic-approved data only.
Open SQLAutomated tests
The test suite checks reproducible generation, public-data contracts and finite model outputs.
Open testsLimitations
Interpretation boundaries
- Public records are synthetic and cannot estimate conditions in Coroico or Bolivia.
- Sixty households are insufficient for complex machine learning or strong subgroup claims.
- Synthetic weights and analytically defined indices are not substitutes for a documented sampling design.
- Cross-sectional associations cannot identify causal effects.
- Multiple imputation and survey-design variance require authorized source data and design variables.
Citation
Citation metadata
Citation metadata are available in the repository's CITATION.cff file. The citation warning states that synthetic results should not be cited as empirical evidence.
Open CITATION.cffAuthor and navigation
Monica Cueto Tapia
Development data analyst portfolio spanning poverty, housing, credit risk, financial development, economic complexity and structural vulnerability.