Figure 1: Standard Queen Logic (Left) vs. ArchipelagoEngine k=5 (Right)
In Spatial Econometrics, standard spatial contiguity models often leave significant portions of island nations mathematically isolated. In the Philippine context, the Queen-Rook approach leaves about 20% of Philippine provinces orphaned, resulting in a fragmented network with only approximately 72% connectivity. This fragmentation introduces systematic predictive bias and significant residual spatial autocorrelation (e.g., Moran's I=0.024, p<0.05 for palay price in the Philippines).
This package implements specialized K-Nearest Neighbor (KNN) logic to bridge these fragmented maritime networks. By enforcing a unified grid (optimized at k=5 using the Philippines as case study), the engine achieves 100% network connectivity and neutralizes spatial bias, enabling robust econometric inference for fragmented geographies.
You may install the package directly to R:
install.packages("ArchipelagoEngine")
The core function, build_archipelago_weight, bridges fragmented networks using optimized KNN logic.
library(ArchipelagoEngine)
library(sf)
library(spdep)
# Load the benchmark map
data(raw_data)
# NOTE: This is a proposed structural baseline that mimics latent maritime infrastructure.
# You can change the k parameter to fit the specific needs, bounds, and theory of your own research.
weights <- build_archipelago_weight(raw_data, k = 5)
# Scans the graph to see if any islands are isolated
# If this returns 1, the computer has verified 100% connectivity
connectivity_status <- spdep::n.comp.nb(weights$neighbours)$nc
print(connectivity_status)
# Plot connectivity
plot(st_geometry(raw_data), border = "lightgrey")
plot(weights$neighbours, st_coordinates(st_centroid(raw_data)),
add = TRUE, col = "#1E90FF", pch = 19, cex = 0.5, lwd = 0.7)
# Result status
mtext(paste("Status: 100% Connectivity Achieved (nc =", connectivity_status, ")"),
side = 1, line = 1, adj = 0.5, cex = 0.9, font = 1, col = "#2C3E50")
Figure 1: Standard Queen Logic (Left) vs. ArchipelagoEngine k=5 (Right)
100% Connectivity via Sparse Matrix Topology: Ensures no island units are mathematically isolated by constructing a regular network graph. This guarantees 100% connectivity while preserving matrix sparsity and preventing global noise.
Automates a listw object: Can be directly feed for spatial regression modeling functions from packages like spatialreg or splm.
Built-In Geographic Data Environment: Ships with a pre-loaded, cleaned raw_data shapefile of the Philippines, mapped to GADM standards.
Neutralizes Spatial Bias: Reduces Global Moran's I to neutralize significant residual spatial autocorrelation.
The package can be scaled to other island nations that face similar matrix inversion failures during their analysis prior to employing Spatial Autoregressive (SAR) or Spatial Error Model (SEM):
Figure 2: Standard Queen Logic (Top) vs. ArchipelagoEngine k=5 (Bottom)
Figure 3: Standard Queen Logic (Left) vs. ArchipelagoEngine k=5 (Right)
A key limitation of ArchipelagoEngine is its reliance on spatial proximity, which may force arbitrary topological connections between islands that lack real-world functional interaction. While this geometric abstraction is an inherent trade-off of the model, integrating transport data — such as the Roll-on/Roll-off (RoRo) networks in the roroph package — offers a more realistic representation of maritime connectivity.
However, researchers must note that RoRo networks can be endogenous to the system, shaped by the very internal economic or geographic factors the model seeks to analyze. Thus, supplementary tests such as Instrumental Variables (IV) must be conducted to control for this potential endogeneity.
The development of ArchipelagoEngine is guided by the #rspatial community to ensure it meets the rigorous standards of spatial econometrics:
sfislands and spdep).roroph package.R in Academia section during the roll-out of issue 2026- W12. Thanks to Jon Calder for going over the submission and Jonathan Carroll for merging the last-minute revision.Anselin, L. (1988). Spatial Econometrics: Methods and Models.
LeSage, J., & Pace, R. K. (2009). Introduction to Spatial Econometrics.
Bivand, R. S., & Wong, D. W. (2018). "Comparing methods for isolating units of spatial autocorrelation." Geographical Analysis.