Table 3. Odds ratios and 95% Bayes CrI from Bayesian geostatistical analysis.
| Model/variable | Coefficient, posterior median (95% Bayes credible interval) | Odds ratio, posterior median (95% Bayes credible interval) |
|---|---|---|
| a) Bayesian geostatistical model, no covariates: | ||
| α (intercept) | 1.51 (-0.42–3.16) | - |
| κ (smoothing parameter) | 0.72 (0.32–1.15) | - |
| ϕ (decay of spatial correlation) | 0.38 (0.11–0.82) | - |
| DIC | 740 | - |
| b) Bayesian geostatistical model with covariates: | ||
| α (intercept) | 2.01 (1.71–2.80) | - |
| Soil moisture | 0.53 (0.25, 0.71) | |
| Daytime lands surface temperature | ||
| ≥ 35°C | 2.81 (2.21, 3.85) | |
| 28−34.9°C | 1.15 (0.98, 2.24) | |
| 24.9−27.9°C | 1.07 (0.59, 1.17) | |
| ≤ 25°C | Reference category | |
| Total edge contrast index | 1.31 (1.27, 2.22) | |
| κ (smoothing parameter) | 0.81 (0.32–1.10) | - |
| ϕ (decay of spatial correlation) | 0.30 (0.11–0.67) | - |
| DIC | 681 | - |
The smoothing parameter, κ in Bayesian spatial models controls the amount of spatial continuity with distance in the S i term. The decay parameter, ϕ refers to the decay of spatial correlation in terms of distance measured in decimal degrees (1 decimal degree at the equator is approximately 120 kilometers). A Bayesian geostatistical model with all covariates selected in the univariate procedure had a DIC value of 684, and the removal of mixed forest and grassland herbaceous cover one at a time resulted in model DIC values of 679 and 681, respectively and were discarded. The removal of Daytime LST, TECI and soil moisture one at a time in that order resulted in DIC values of 694, 691 and 703, respectively and were retained in the final model. Age and sex covariates were inserted in the Bayesian geostatistical model with covariates, one at a time, and ≥10% changes to posterior median estimates was not noted following insertions. The ‘unknown’ groups for age and sex were used as reference categories.