Crowded but not poorest: addressing child stunting beyond wealth quintiles in urban Bangladesh

Published: 13 October 2026| Version 3 | DOI: 10.17632/y7zpd4gnt6.3
Contributors:
,

Description

This deposit contains the analysis code and a complete set of run outputs for a study of multidimensional housing deprivation and child undernutrition among urban children in Bangladesh, using the 2022 Bangladesh Demographic and Health Survey (BDHS). The study develops an empirically derived latent-class typology of household housing/WASH deprivation and examines whether it provides information about child stunting and socioeconomic inequality beyond the conventional DHS household wealth index. Contents 1. full_analysis_poLCA_ALL_IN_ONE.R — a single, self-contained R script reproducing the full analysis: latent class analysis of six housing/WASH indicators (poLCA), survey-weighted prevalence estimation, multilevel logistic regression, a formal wealth-only versus wealth-plus-housing-class model comparison, Erreygers concentration-index decomposition of socioeconomic inequality, sensitivity analyses (alternative crowding thresholds; alternative numbers of latent classes), adjustment for latent-class classification uncertainty (posterior-probability weighting and pseudo-class multiple imputation), and complex-survey-weighted (svyglm) analyses. 2. README.md and Data/README_DATA_ACCESS.md instructions for running the script and for obtaining the underlying BDHS 2022 microdata. 3. Data/data_dictionary_bdhs2022_urban_analysis_file.csv a variable-by-variable dictionary for the script's required input file. 4. bdhs_r_analysis_poLCA_output_<timestamp>/ the full output of one executed run: stage-by-stage CSV results (data-integrity checks, latent class fit statistics and class profiles, survey-weighted prevalence, regression coefficients, concentration-index decomposition, sensitivity-analysis results, targeting-performance comparisons), the fitted latent class model object, four figures (TIFF, 300 dpi), and the four supporting-information files (S1 Table, S2 Table, S3 Table, S1 Text; .docx), each generated directly by the script not transcribed by hand.

Files

Steps to reproduce

1. Request access to the Bangladesh 2022 DHS from The DHS Program (free registration): https://dhsprogram.com/data/dataset_admin/login_main.cfm — download the Children's Recode (BDKR81FL) and Household Recode (BDHR81FL). 2. Restrict each recode to urban clusters (V025 == 1 in KR; HV025 == 1 in HR), construct the variables listed in `Data/data_dictionary_bdhs2022_urban_analysis_file.csv`, and left-join KR onto HR by cluster + household ID to produce `bdhs2022_urban_analysis_file.csv`. (The authors' own derivation script, `build_analysis_file.R`, performs this step and is available from the corresponding author on request — it is not included in this deposit because it operates on restricted-access DHS files.) 3. Place `bdhs2022_urban_analysis_file.csv` in the same folder as `full_analysis_poLCA_ALL_IN_ONE.R`. 4. Open the script in RStudio and click Source, or run `Rscript full_analysis_poLCA_ALL_IN_ONE.R` from a terminal. R >= 4.2 required (developed/tested on R 4.3.3); missing packages (poLCA, dplyr, lme4, survey, car, officer, flextable) install automatically on first run. 5. Runtime is several minutes. On completion two folders appear next to the script: `bdhs_r_analysis_poLCA_output_<timestamp>/` (every stage's raw CSV, the fitted latent class model, and a `Figures/` subfolder with Fig1-Fig4) and `Supplementary_Tables/` (S1 Table, S2 Table, S3 Table, S1 Text as .docx). 6. To check reproduction against the manuscript, compare `stage1_lca_fit_table.csv` (latent class fit statistics), `stage1_class_profile.csv` (class-conditional item probabilities for the selected solution), and `SUMMARY_crowded_deprived_OR_across_specs.csv` (the Crowded Deprived odds ratio across all nine specifications) against the corresponding manuscript tables/figures. Note: latent class analysis fits by multi-start EM. A rerun is not guaranteed to be bit-identical to the deposited run even with the same seed, though the solution has been cross-checked across independent implementations (see `Supplementary_Tables/S1_Text.docx`, generated by this run, for exact package/version provenance and the seed's scope).

Institutions

Categories

Health Policy, Public Health

Licence