First release of the R package, a port of the Python package rd2dnet 0.1.0 with the
same methods and, on the bundled mock city, the same numbers.
- Network boundary discontinuity estimation (
rdnet()): side-specific local polynomials in network coordinates, robust bias-corrected inference, uniform confidence bands, fuzzy designs, covariates, derivative and directional-derivative targets, HC0-HC3, cluster and network/Euclidean Conley standard errors. - Aggregate effects as in rd2d:
wbate()andlbate(), with the joint covariance across evaluation points stored in the result (vcov()). - Bandwidth selection (
rdbwnet()), ported from rd2d's direct plug-in, including the coverage-error-optimal variants. - Density (sorting) test (
density()method for network coordinates), ported from rddensity, with a street-length normalisation. - User-defined network metrics: edge costs, travel time, node contrasts, transition
penalties between road classes (
transition_cost()) and precomputed matrices. - Networks from data frames, adjacency matrices, igraph graphs or OpenStreetMap (via dodgr), with sf geometries accepted for points, boundaries, regions and masks.
- Shortest paths, nearest-segment snapping and the sparse Conley kernel are implemented in C++.
- Vignette with donut, covariate-balance, placebo-boundary and difference-in-discontinuities recipes.