NEWS

---

rd2dnet 0.1.0 (2026-10-28)

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() and lbate(), 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.