Tools for Grid-Based Survey Sampling Design
Multi-stage cluster surveys of households are commonly performed by
governments and programmes to monitor population-level demographic, social,
economic, and health outcomes. Generally, communities are sampled from
subpopulations (strata) in a first stage, and then households are listed and
sampled in a second stage. In this typical two-stage design, sampled
communities are the Primary Sampling Units (PSUs) and households are the
Secondary Sampling Units (SSUs). Census data typically serve as the sample
frame from which PSUs are selected. However, if census data are outdated
inaccurate, or too geographically course, gridded population data (such as
< http://www.worldpop.org.uk>) can be used as a sample frame instead.
GridSample (<10.1186>) generates PSUs from
gridded population data according to user-specified complex survey design
characteristics and household sample size. In gridded population sampling,
like census sampling, PSUs are selected within each stratum using a
serpentine sampling method, and can be oversampled in urban or rural areas to
ensure a minimum sample size in each of these important sub-domains.
Furthermore, because grid cells are uniform in size and shape, gridded
population sampling allows for samples to be representative of both the
population and of space, which is not possible with a census sample frame.10.1186>
Grid-based alternative to doing DHS-style survey sample design. An open source R package developed as a WorldPop project.
gridsample v0.2.1 (Update date: 2017-04-05)
- Fix cfg_desired_cell_size and aligns with paper (https://doi.org/10.1186/s12942-017-0098-4)
- Implements cfg_random_number as specified in paper (https://doi.org/10.1186/s12942-017-0098-4).
- Modifies file paths for interoperability in vignette.
gridsample v0.2.0 (Update date: 2017-04-05)
- Added Rwanda 2010 DHS vignette
- Added cfg_psu_growth parameter to enable or disable PSU growth
- Fixed incorrect usage of cfg_max_psu_size
- Additional documentation and example in gs_sample()