Statistical Methods
Check the Domain Before Interpolating on a Triangulation
Define the task
A Delaunay triangulation partitions a point set into simplices for geometric operations such as local interpolation. It does not make scattered observations representative of an unsampled region. Define coordinate units, scaling, duplicate handling and the domain in which interpolation is scientifically defensible. Geometric proximity can change substantially after rescaling an axis. A successful triangulation is therefore only one part of a defensible interpolation workflow.
Carry out the check
Inspect degenerate or nearly collinear inputs and read the geometry library's diagnostics. Locate each query point in a simplex before using barycentric coordinates, and handle points outside the triangulated region explicitly. Do not interpret a negative simplex index as a valid last element of an array. Preserve the vertex-to-data mapping and inspect any omitted or coplanar points. A geometric triangle crossing a physical barrier may be mathematically valid but scientifically inappropriate for interpolation.
Worked hypothetical example
For a hypothetical triangle with vertices (0,0),(1,0),(0,1), the point (0.25,0.25) has barycentric weights (0.5,0.25,0.25). Vertex values (0,4,8) interpolate to three. The point (1,1) lies outside this triangle and needs an explicit extrapolation policy. These hand calculations do not establish the quality of a spatial model. Keep a test with known weights and a known outside point to verify indexing, domain checks and reported missing outputs.
Method checked 21 November 2026 against Primary method documentation: Delaunay.