\begin{table}[H]
\centering
\begin{threeparttable}
\caption{Strict spatial-screen design and sensitivity diagnostics}
\label{tab:strictspatialdiagnostics}
\small
\setlength{\tabcolsep}{3.8pt}
\begin{tabularx}{0.96\textwidth}{>{\raggedright\arraybackslash}p{0.31\textwidth}c>{\raggedright\arraybackslash}X}
\toprule
Diagnostic & Value & Interpretation \\
\midrule
Coordinate-usable main sample & 9 routes & All main-sample ranks and flags; each endpoint has prototype-usable coordinates. \\
Manual-coordinate feasibility layer & 13 candidates & Shown only as a map/context layer pending manual endpoint verification. \\
Geometry grid & 48 settings & Four buffers, four coordinate-jitter scales, three pseudocounts, and 200 fixed-seed draws each. \\
Population-model rank check & 0.817 & Spearman correlation of WorldPop 2020 and GHSL 2025 projected-population ranks. \\
Population top-three overlap & 3/3 & Same three strict routes in the two population top-three screens. \\
Geometry-grid minimum overlap & 1/3 / 3/3 & Minimum absolute / relative top-three overlap with the 500 m baseline. \\
POI-definition grid & 7 settings & Baseline, three leave-one-sensitive-family-out definitions, and three alternative service-opportunity denominators. \\
POI-definition minimum overlap & 2/3 / 2/3 & Minimum absolute / relative top-three overlap with the baseline POI definition. \\
\bottomrule
\end{tabularx}
\begin{tablenotes}
\footnotesize
\item Notes: WorldPop 2020 is the baseline count layer. GHSL 2025 is an alternative projected residential-population model used only for rank sensitivity; agreement is not independent validation. Public endpoint lines, population, and OSM facilities are prospective screening inputs, not filed trajectories, acoustic dose, payment receipt, or welfare outcomes.
\end{tablenotes}
\end{threeparttable}
\end{table}