GeoDevelopment Observatory of Bolivia / Tools
expdpy: explore and model panel data
A companion package for specialised panel-data methods, beyond geo4bolivia’s built-in analyses: the split of variation between and within units, correlated random effects, instrumental variables, event studies and robust inference, each with a plain-language reading. geo4bolivia hands it Bolivian panels in one line.
geo4bolivia runs the fundamental panel analyses itself: trends, regressions, and pooled, between, fixed- and random-effects models with a Hausman test. expdpy (a Python port of ExPanDaR) is its companion for going further with panels. Its explore functions describe a panel’s structure, decompose variation into between and within components (xtsum), and chart trajectories, distributions, transitions, missing values and correlations. Its analyze functions add correlated random effects, joint tests, fixest-style regressions with fixed effects and clustered or Driscoll–Kraay standard errors, wild-bootstrap inference, instrumental variables and event studies. Its learn functions simulate each concept from a known truth. Every result has an interactive figure or table and an interpret() method. The observatory’s datasets are exactly the long panels expdpy expects: geo.for_expdpy() hands them over, labelled and with the department of each municipality, in one line. The SDG Atlas gives two periods (~2012 and ~2024) for 343 units; night-time lights give a nine-year panel (2012–2020) for 333.