A spatio-temporal approach to geographically and temporally weighted regression with fixed and adaptive kernel bandwidth selection based on corrected generalized cross-validation

Bandwidth selection plays a decisive role in Geographically and Temporally Weighted Regression (GTWR) by controlling local smoothing and influencing model accuracy, complexity, and stability. This study evaluates the corrected generalized cross-value (cGCV) as a bandwidth-selection criterion for GTWR and compares kernel functions and bandwidth configurations in modeling the prevalence of childhood stunting in Indonesia. The analysis consisted of two stages. First, Monte Carlo simulations compared cross-validation (CV), generalized cross-validation (GCV), and cGCV under weak, moderate, and strong spatio-temporal heterogeneity. Each scenario involved 50 replications with 120 observations. Performance was assessed using adjusted R2 , root mean square error (RMSE), and corrected Akaike information criterion (AICc). Second, the selected criterion was applied to provincial stunting data from 2020 to 2024, evaluating four kernel functions under fixed and adaptive bandwidth specifications. The simulation results consistently favored cGCV in all scenarios. The adjusted R2 values were 0.911, 0.709 and 0.491, while the RMSE values were 0.386, 1.796 and 3.920, respectively. Corresponding AICc values were 262.090, 563.516, and 698.033. In the empirical analysis, the fixed exponential kernel achieved the lowest cGCV value of 6.587, with a spatio-temporal bandwidth of 15.393. The resulting GTWR model achieved an R2 of 0.906, adjusted R2 of 0.902, RMSE of 1.744, and AICc of 1228.978, outperforming ordinary least squares (OLS), temporally weighted regression (TWR) and geographically weighted regression (GWR). These findings indicate that cGCV offers a practical bandwidth-selection approach under the simulated conditions. At the same time, GTWR captures provincial and temporal variations in factors associated with stunting, supporting locally tailored public health interventions.

Authors

Institutions

Publication Details

Journal
Hacettepe Journal of Mathematics and Statistics
Published
2026-11-16
DOI
https://doi.org/10.15672/hujms.1838917
Primary Topic
Spatial and Panel Data Analysis
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A spatio-temporal approach to geographically and temporally weighted regression with fixed and adaptive kernel bandwidth selection based on corrected generalized cross-validation

Robiansyah Putra, Fidelis Nofertinus Zai
Hacettepe Journal of Mathematics and Statistics
Spatial and Panel Data Analysis
article

A spatio-temporal approach to geographically and temporally weighted regression with fixed and adaptive kernel bandwidth selection based on corrected generalized cross-validation

Robiansyah Putra, Fidelis Nofertinus Zai
article en

Abstract

Bandwidth selection plays a decisive role in Geographically and Temporally Weighted Regression (GTWR) by controlling local smoothing and influencing model accuracy, complexity, and stability. This study evaluates the corrected generalized cross-value (cGCV) as a bandwidth-selection criterion for GTWR and compares kernel functions and bandwidth configurations in modeling the prevalence of childhood stunting in Indonesia. The analysis consisted of two stages. First, Monte Carlo simulations compared cross-validation (CV), generalized cross-validation (GCV), and cGCV under weak, moderate, and strong spatio-temporal heterogeneity. Each scenario involved 50 replications with 120 observations. Performance was assessed using adjusted R2 , root mean square error (RMSE), and corrected Akaike information criterion (AICc). Second, the selected criterion was applied to provincial stunting data from 2020 to 2024, evaluating four kernel functions under fixed and adaptive bandwidth specifications. The simulation results consistently favored cGCV in all scenarios. The adjusted R2 values were 0.911, 0.709 and 0.491, while the RMSE values were 0.386, 1.796 and 3.920, respectively. Corresponding AICc values were 262.090, 563.516, and 698.033. In the empirical analysis, the fixed exponential kernel achieved the lowest cGCV value of 6.587, with a spatio-temporal bandwidth of 15.393. The resulting GTWR model achieved an R2 of 0.906, adjusted R2 of 0.902, RMSE of 1.744, and AICc of 1228.978, outperforming ordinary least squares (OLS), temporally weighted regression (TWR) and geographically weighted regression (GWR). These findings indicate that cGCV offers a practical bandwidth-selection approach under the simulated conditions. At the same time, GTWR captures provincial and temporal variations in factors associated with stunting, supporting locally tailored public health interventions.

Hacettepe Journal of Mathematics and Statistics(Advanced Online Publication)
Universitas Sumatera Utara (ID)
Universitas Sumatera Utara
Spatial and Panel Data Analysis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon