Data & Methodology — Bangladesh DHS Health Atlas

⚠ Why does the dashboard say “Sample data”?

The badge in the top bar shows “Sample data” (amber) until the real survey numbers are loaded. In that mode the values you see are approximate placeholders generated only so the interface works and can be demonstrated — they are not real BDHS results and must not be quoted. Every placeholder is marked with a grey/red reliability dot for the same reason.

The badge turns green “Live data” automatically once real, survey-weighted estimates are loaded into data/master.json. There are two honest ways to produce those real numbers, and neither one invents or predicts anything:

Process the survey files. Run scripts/convert_dta.py on the original BDHS Stata .dta files. It applies the official sampling weights and computes each indicator directly from the raw survey responses. No Python on your PC? Run it free in Google Colab (see the README) and upload the resulting master.json to Hostinger.
Transcribe the published report. The figures in the BDHS 2022 final report (FR386.pdf) are the official weighted estimates; those can be entered directly.

Until then, treat the dashboard as a working demo, not a source of numbers.

1. What data powers this atlas

The atlas draws on the Bangladesh Demographic and Health Surveys (BDHS) — large, nationally representative household surveys implemented by the National Institute of Population Research and Training (NIPORT) with technical support from The DHS Program (ICF) and funding from USAID. Bangladesh has run BDHS rounds in 1993-94, 1996-97, 1999-2000, 2004, 2007, 2011, 2014, 2017-18 and 2022. This project uses the rounds you supplied, from 1996-97 through 2022.

Each round interviews tens of thousands of households and eligible women (and, in some rounds, men), collecting fertility, family-planning, maternal and child-health, nutrition, anaemia, mortality and household-environment information using standardised questionnaires that are comparable across countries and over time.

2. The survey rounds in this project

Round	Folder	Main topics added / notes
1996-97	BD_1996-97_DHS…	Core fertility, FP, child health
1999-00	BD_1999-00_DHS…	+ nutrition/anthropometry expansion
2004	BD_2004_DHS…	+ HIV/AIDS knowledge, wider health
2007	BD_2007_DHS…	Maternal & newborn care detail
2011	BD_2011_DHS…	+ biomarkers (anaemia, BP)
2014	BD_2014_DHS…	Rangpur reported as a division
2017-18	BD_2017-18_DHS…	Mymensingh added; biomarkers
2022	BD_2022_DHS…	Largest round; first designed for some district-level estimates

Exact sample sizes and the indicators available differ by round — always confirm against the corresponding final report before quoting a figure.

3. The data files (recodes) explained

DHS distributes each round as several standardised “recode” Stata files. Each file is a different unit of analysis. Inside every round folder you supplied you will see codes like BDIR…, BDKR… etc.:

Code	Recode	One row =	Used here for
IR	Individual / Women	a woman 15–49	Contraception, ANC, skilled birth, fertility, child marriage, women's anaemia/BMI, education
MR	Men	a man 15–54/59	Men's knowledge/behaviour (optional)
KR	Children (Kids)	a living child < 5	Immunisation, diarrhoea/ARI treatment, breastfeeding, child anthropometry
BR	Births	a birth in the birth history	Infant & under-5 mortality rates
HR	Household	a household	Water, sanitation, electricity, wealth quintile
PR	Household members (Persons)	a person in the household	Anthropometry & anaemia rosters, education, residence
CR	Couples	a matched couple	Couple-level FP analysis (optional)
HW / WI	Height-Weight / Wealth Index	child / household	Older-round nutrition & wealth
GC	Geographic Covariates	a survey cluster	Cluster context (see the PDF descriptions in each GC folder)

scripts/convert_dta.py opens the correct recode for each indicator, applies the weight, groups by division/district, and writes the results the dashboard reads.

4. How each indicator is defined

The atlas computes 24 indicators across four themes. Each is a survey-weighted proportion (or a rate/mean) using the standard DHS variable shown:

Theme	Indicators (DHS variable)
Child health & mortality	Stunting (hw70<−2SD), wasting (hw72), underweight (hw71), under-5 & infant mortality (birth history), Penta-3 (h7), measles (h9), full immunisation, ORS for diarrhoea (h13)
Maternal & reproductive	Modern contraception (v313=3), unmet need (v624), ANC 4+ (m14), skilled birth attendance (m3), facility delivery (m15), total fertility rate, married before 18 (v511)
Nutrition & anaemia	Child anaemia, women's anaemia (v457), exclusive breastfeeding, women thin BMI<18.5 (v445)
Household, WASH & wealth	Improved water (hv201), improved sanitation (hv205), electricity (hv206), women with secondary+ education (v106)

Full definitions appear in the tooltip on each indicator button and in data/indicators.json.

5. Weighting, wealth index & representativeness

Sampling weights

BDHS uses a stratified, two-stage cluster sample, so every record carries a sampling weight (v005/hv005, divided by 1,000,000). All estimates here are weighted so they represent the population, not just the sample.

Wealth quintiles

The DHS wealth index (v190/hv270) is built from household assets and housing characteristics and split into five equal groups — poorest to richest — used by the Wealth filter.

Geographic representativeness — important

BDHS is designed to be statistically representative at the division level for every round, and at the district level only for the 2022 round (and partially 2017-18). It is not representative at the upazila level in any round. The atlas therefore attaches a reliability flag to every value:

Flag	Meaning
Reliable	Adequate sample (n ≥ 200) — usually division cells
Indicative	Small sample (50–199) — read with caution
Not representative	Very small sample (< 50)
Modelled / sample	Placeholder or estimated from the parent area (e.g. all upazilas, and everything while in “Sample data” mode)

Upazila colours are shown for visual exploration only — do not use them for official reporting.

6. Getting the real numbers in (recap)

Run scripts/convert_dta.py on the BDHS .dta files — locally or in free Google Colab (no install). It outputs a real master.json.
In Hostinger File Manager, replace public_html/data/master.json with it.
Reload — the badge turns green “Live data” and every figure is a real weighted BDHS estimate.

Alternatively the official figures in the BDHS 2022 report (FR386.pdf) can be transcribed directly for the 2022 round.

Source: Bangladesh Demographic and Health Surveys — NIPORT, Ministry of Health & Family Welfare, and The DHS Program (ICF). This atlas is an independent visualisation.