debiasRdata

debiasRdata is the optional empirical data companion package for debiasR. It keeps empirical travel-to-work data out of the main methods package, so debiasR can stay small, method-focused, and CRAN-friendly.

The package supplies documented data assets only. Bias adjustment, modelling, validation, and example-data loading logic live in debiasR.

Contributions are made through pull requests. See CONTRIBUTING.md for the review workflow and data-release checks.

Installation

Install the package from GitHub with pak:

pak::pak("de-bias/debiasRdata")

If you do not use pak, install with remotes instead:

remotes::install_github("de-bias/debiasRdata")

Included datasets

Dataset names identify the geography explicitly: msoa_... for MSOA assets and lad_... for LAD/LTLA assets, with a census_ prefix for Census benchmarks.

msoa_OD_travel2work contains observed mobile-phone-derived MSOA travel-to-work OD flows from Zenodo record https://doi.org/10.5281/zenodo.13327082.

census_msoa_OD_travel2work contains the matching Census 2021 ODWP01EW MSOA workplace-flow benchmark from the Office for National Statistics via Nomis.

lad_OD_travel2work contains the same observed mobile-phone-derived flows aggregated to LAD22 origin-destination pairs with the ONS OA_to_LSOA_to_MSOA_to_LAD_Dec_2021_EW_V3.csv lookup.

census_lad_OD_travel2work contains the Census 2021 ODWP01EW LTLA workplace-flow benchmark. The observed LAD aggregate uses LAD22CD from the ONS lookup; the Census LAD/LTLA benchmark uses the authority codes supplied directly by ODWP01EW_LTLA.csv.

lad_centroids contains ONS LAD December 2021 representative coordinates. It lets debiasR compute selected-area LAD distances without packaging a full OD distance matrix.

msoa_centroids contains MSOA representative coordinates. It lets debiasR compute selected-area MSOA distances without packaging a full OD distance matrix.

msoa_covariates and lad_covariates contain selected Census 2021 topic-summary covariates for MSOA and LAD/LTLA areas.

coverage_lad contains LAD/LTLA benchmark population counts and mobile-phone- derived active-user counts for coverage-bias examples.

coverage_msoa contains the corresponding MSOA benchmark population counts and mobile-phone-derived active-user counts for coverage-bias examples.

The four OD datasets use the same normalized schema:

Column	Type	Meaning
origin	character	Origin MSOA or LAD/LTLA code
destination	character	Destination MSOA or LAD/LTLA code
flow	numeric	Non-negative OD flow

The two covariate datasets use this schema:

Column	Type	Meaning
area	character	MSOA or LAD/LTLA code
name	character	Area name
year	integer	Census year
per_ukborn	numeric	Percentage UK-born
per_age_20.29	numeric	Percentage aged 20-29
per_age_70plus	numeric	Percentage aged 70 or over
per_level4	numeric	Percentage with level 4 qualifications
per_hh_no_centralheat	numeric	Percentage of households without central heating
per_NS_SeC_L13_routine	numeric	Percentage in NS-SeC L13 routine occupations
rural_pct	numeric	Percentage rural

The coverage datasets use this schema:

Column	Type	Meaning
date	integer	Reference year
name	character	MSOA or LAD/LTLA name
code	character	MSOA or LAD/LTLA code
population	integer	Benchmark population count
user_count	numeric	Mobile-phone-derived active-user count

Usage

library(debiasRdata)

head(msoa_OD_travel2work)
head(census_msoa_OD_travel2work)
head(lad_OD_travel2work)
head(census_lad_OD_travel2work)
head(lad_centroids)
head(msoa_centroids)
head(msoa_covariates)
head(lad_covariates)
head(coverage_lad)
head(coverage_msoa)

debiasR consumes these objects conditionally:

library(debiasR)

ex <- debiasR_example_data(n_areas = 25)
names(ex)

Compressed normalized CSV files are also installed under inst/extdata and can be located with:

debiasRdata_path("msoa_OD_travel2work")
debiasRdata_path("census_msoa_OD_travel2work")
debiasRdata_path("lad_OD_travel2work")
debiasRdata_path("census_lad_OD_travel2work")
debiasRdata_path("lad_centroids")
debiasRdata_path("msoa_centroids")
debiasRdata_path("msoa_covariates")
debiasRdata_path("lad_covariates")
debiasRdata_path("coverage_lad")
debiasRdata_path("coverage_msoa")

The helper only locates installed files. It does not download data during package use.

Sources and licenses

The mobile-phone-derived table is derived from:

• Title: Anonymised human location data for urban mobility research
• DOI: https://doi.org/10.5281/zenodo.13327082
• Source file: msoa_OD_travel2work.csv.gz
• License: Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/
• Packaged modifications: columns were normalized to origin, destination, and flow; county columns were removed; OD pairs were validated and aggregated where needed.

The MSOA Census benchmark is derived from:

• Table: Census 2021 ODWP01EW, MSOA workplace-flow origin-destination table
• Source institution: Office for National Statistics, accessed via Nomis
• License: Open Government Licence v3.0, https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/
• Packaged modifications: rows were filtered to people working in the UK but not working at or from home; non-MSOA destination codes were removed; columns were normalized to origin, destination, and flow.

The LAD-level observed table is derived from the Zenodo MSOA travel-to-work table using the ONS OA_to_LSOA_to_MSOA_to_LAD_Dec_2021_EW_V3.csv lookup. The lookup is published by ONS under the Open Government Licence v3.0.

The LAD/LTLA Census benchmark is derived from:

• Table: Census 2021 ODWP01EW, LTLA workplace-flow origin-destination table
• Source file: ODWP01EW_LTLA.csv in the Nomis Census 2021 origin-destination zip
• Source institution: Office for National Statistics, accessed via Nomis
• License: Open Government Licence v3.0, https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/
• Packaged modifications: rows were filtered to people working in the UK but not working at or from home; columns were normalized to origin, destination, and flow; valid England/Wales local authority codes were retained.

The LAD centroid table is derived from:

• Boundary: ONS Local Authority Districts December 2021 GB boundary file LAD_Dec_2021_GB_BFC_2022.gpkg
• Source institution: Office for National Statistics
• License: Open Government Licence v3.0, https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/
• Packaged modifications: LAD code, LAD name, British National Grid easting / northing, and longitude / latitude fields were retained and renamed to area, name, easting, northing, longitude, and latitude.

The MSOA centroid table is derived from the local DEBIAS geography centroid input msoa_centroids.rds, available under the Open Government Licence v3.0. The packaged table keeps MSOA code, MSOA name, British National Grid easting / northing, and longitude / latitude fields under the same schema as lad_centroids.

The covariate tables are derived from:

• Collection: Census 2021 topic-summary inputs
• Source files: combined-data-msoa.csv, combined-data-lad.csv, and dictionary.csv
• Source institution: Office for National Statistics
• License: Open Government Licence v3.0, https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/
• Packaged modifications: geography.code was renamed to area, geography to name, and date to year; the seven requested label-5 covariates were retained; source column per_age_20-29 was renamed to the label-5/R-facing name per_age_20.29.

The LAD coverage table is derived from the same Zenodo mobile-phone-derived data release used for the packaged observed travel-to-work OD assets, licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). The packaged table keeps benchmark population counts, renames the LAD-level aggregate active-user count to user_count, sets date to 2021, and maps LAD/LTLA codes from lad_covariates by area name. It contains LAD/LTLA-level aggregate counts only; it does not contain individual records, device identifiers, raw trajectories, or raw mobile-phone data.

The MSOA coverage table is derived from the local DEBIAS coverage_msoa.rda active-population-bias output. The active-user counts are derived from the same Zenodo mobile-phone-derived data release used for the packaged observed travel-to-work OD assets, licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Benchmark population counts are ONS-derived under the Open Government Licence v3.0. The packaged table keeps date, MSOA name, benchmark population, and aggregate active-user user_count, renames MSOA21CD to code, and removes rows without valid MSOA codes. It contains MSOA-level aggregate counts only; it does not contain individual records, device identifiers, raw trajectories, or raw mobile-phone data.

Build metadata, checksums, row counts, and transformation notes are recorded in inst/metadata/source-metadata.json.

Planned data

Full OD distance tables are not packaged. For LAD examples, debiasR computes selected-area distances from lad_centroids. For MSOA examples, debiasR can compute selected-area distances from msoa_centroids.