Household survey data and analysis code for a study of overlapping household differentiation in Kombissiri, Burkina Faso

This deposit contains the anonymised household-survey dataset and the complete
analysis code underpinning the article “Overlapping household differentiation
and its spatial expression in the Sahelian small town of Kombissiri, Burkina Faso”
(Zorom et al., submitted to Scientific Reports). A live, deployed version of the
full analysis is available at https://mzorom.shinyapps.io/Typo_Kombissiri_BF/.
The study surveyed 268 households in Kombissiri (Bazèga Province, Burkina Faso)
between April and June 2024 as part of the RurbanClimate project (DANIDA
grant 20-11-KU), and describes how households differ along socio-economic,
demographic and land-tenure dimensions using Factor Analysis of Mixed Data
(FAMD) on 45 active variables (the administrative sector is retained as an active
variable; only the GPS coordinates are excluded from the factor space), Ward
clustering, and spatial autocorrelation analysis.
The material is deposited so that all non-spatial results can be reproduced from
the public data and the complete analysis, including spatial outputs, can be
reproduced when the restricted georeferenced dataset is supplied to the same
pipeline. The central empirical result of the study is a negative one: household
differentiation in Kombissiri is strongly overlapping and only weakly separated
rather than organised into discrete types, and the deposited code makes the
diagnostics behind that conclusion (silhouette widths, cluster-number indices,
bootstrap stability, effect sizes, and spatial autocorrelation) fully inspectable
and reproducible.
Two things are worth stating up front for anyone reusing the data. First, the
survey used a multi-stage cluster design stratified by administrative sector, so
inferential statistics computed under a simple-random-sampling assumption are
indicative rather than exact. Second, the household-level geographic coordinates
are sensitive and have been removed from this public deposit to protect respon
dent confidentiality. As a result the deposit reproduces all of the non-spatial
results; the spatial figures, the global Moran’s I tests and the LISA maps require
the georeferenced version, which is available from the corresponding author under
the data-use terms described below.