| A Non-confounded Spatio–Temporal Count Model for Analysis of Wet-Day Frequency Data |
| کد مقاله : 1034-SPATIAL (R1) |
| نویسندگان |
|
مهسا نادی فر1، حسین باغیشنی *2 1Department of Statistics | University of Pretoria 2دانشگاه صنعتی شاهرود |
| چکیده مقاله |
| We present a non-confounded spatio–temporal count regression based on a Gamma–Count (GC) model for spatially distributed counts with dispersion, latent dependence, and spatial confounding. Within a structured additive regression, we combine thin‐plate splines, a Matérn SPDE field with an AR(1) time effect, and optional nonlinear effects. Inference is conducted within a Bayesian framework using the integrated nested Laplace approximation (INLA) method. The proposed model, applied to the monthly wet-day frequency at 94 Quebec stations from January to March 2025 with unequal exposure, outperforms the relevant Poisson and negative binomial baselines, showing a strong positive effect of total precipitation and negligible temperature effects after deconfounding. Posteriors indicate under‐dispersion, long spatial range, high temporal persistence, and calibrated latent‐field maps. The framework advances Bayesian analysis of environmental counts by jointly addressing dispersion, dependence, and confounding, and offers a practical template for hydrologic and climate applications. |
| کلیدواژه ها |
| Gamma-count model, Spatial+, Confounding, Spatio-temporal data, Precipitation, SDG. |
| وضعیت: پذیرفته شده برای ارائه شفاهی |