Companion to the journal
Data Darbar
Additions and corrections
Adaad's charts are built on this explorer, and it keeps changing. A reader
who checks a figure some months after it was published deserves to know
whether the layer behind it has since been rebuilt, so every addition, every
correction and every change of default is dated here. Corrections to what
Adaad itself published are a different record and live in the
corrections log.
9 additions 3 corrections 4 changes
The layer behind “The unequal road to care” and its method notebook, frozen 9 September: modelled travel time to the nearest mapped health facility on the Malaria Atlas motorised and walking-only surfaces, weighted by WorldPop 2020. For 147 districts it carries medians, means and the shares beyond 30, 60 and 120 minutes; for 553 tehsils, means and shares as published. It is on the map under Health, and in the warehouse as health_access_district and health_access_tehsil with worked console examples. A README states what the measure is and is not: potential travel to the nearest facility that exists, not an observed journey and not a statement about the care available there.
The school layer built for “How far is the girls’ school?” is now Darbar’s. The warehouse holds schools_pk — 125,317 government schools, one row each, carrying how every coordinate was arrived at so nobody reads a tehsil centroid as a GPS fix — beside district distances and enrolment by sex, the layer’s own coverage table, and the Mouza Census checks re-run on the released version. The map gains a tehsil group under Education: population-weighted median distance to the nearest government school of each sex at three levels, the girls-minus-boys gap, and the share of people more than 5 km from a girls’ middle school, for 498 of 553 tehsils. The district aggregates of this run reproduce the piece to 0.001 km.
The Poverty & Wealth page is a topic on the map instead: the PSLM 2019-20 multidimensional poverty index by district, and Meta’s relative wealth, WorldPop population and VIIRS night-time lights by tehsil. Choosing a tehsil dataset switches the geometry, and night lights get a 2020–2026 slider in place of the census-year buttons. The seven Karachi districts show the city-wide index, flagged as such, and withheld units — small samples, uninhabited tehsils — keep their own fill and note rather than being coloured as though they had a value. The map reads its topic, group and indicator from the URL, so one view can be linked to.
DuckDB compiled to WebAssembly, querying static Parquet in the browser: nothing is uploaded and there is no server to ask permission from. It now holds 33 SBP EasyData datasets, 306,875 observations, beside a catalogue of all 226 the bank publishes. Any table can be pulled whole rather than queried, and a download can carry its method notes: a README with the table's source, unit, caveats and column dictionary. The dictionary is also a page of its own. Units differ series by series, and the remittance-by-country series nest inside one another, so adding them all overstates the total by 42 per cent; the catalogue spells out the partition that reconciles.
