Bayesian Hot Spot Analysis
Also known as: Bayesian spatial cluster detection, Bayesian disease mapping hot spots, empirical Bayesian hot spot analysis, Bayesian spatial smoothing hot spots
Bayesian Hot Spot Analysis identifies spatial clusters of elevated risk or intensity by combining observed data with prior beliefs about spatial structure. It uses Bayesian smoothing — pooling information across neighboring areas — to stabilize estimates in small areas and then flags locations where the posterior probability of exceeding a risk threshold is high.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Bayesian hot spot analysis when you have areal count or rate data (disease cases, crime incidents, accident counts) distributed across areas with highly variable populations or exposure denominators, and small-area reliability is a concern. It is the preferred approach in disease mapping, epidemiological surveillance, and public health planning. It is particularly valuable when areas differ greatly in size or population, making raw rates unstable. Do not use it as a drop-in replacement for frequentist Getis-Ord when data are point patterns without denominators or when the spatial weight structure is unclear; also avoid it if computational resources or Bayesian modeling expertise are unavailable, since MCMC or INLA fitting is required.
Strengths & limitations
- Posterior probability of being a hot spot provides direct, interpretable uncertainty quantification — unlike p-values from frequentist methods.
- Bayesian spatial smoothing reduces false positives from small-number instability in sparsely populated areas.
- The BYM and related models separate spatially structured clustering from unstructured overdispersion, revealing the nature of clustering.
- Prior information (e.g., known risk factors, historical rates) can be formally incorporated.
- Hierarchical extensions naturally handle missing data, covariate adjustment, and multi-level structures.
- Computationally intensive: MCMC sampling or INLA approximation is required, adding complexity compared to classical methods.
- Results depend on the choice of prior and spatial weight matrix (e.g., queen contiguity vs. distance-based); sensitivity analysis is needed.
- Requires expertise in Bayesian modeling and spatial statistics to specify, run, and interpret models correctly.
- Over-smoothing can mask genuinely sharp boundaries between hot and cold spots if the spatial prior is too strong.
Frequently asked
How does Bayesian hot spot analysis differ from Getis-Ord Gi*?
Getis-Ord Gi* is a frequentist test that computes a z-score based on local weighted averages and flags extreme values. Bayesian hot spot analysis models the full posterior distribution of relative risk, borrows strength from neighbors via spatial priors, and reports the probability that each area truly exceeds a risk threshold. It is more robust to small-number instability and provides explicit uncertainty quantification.
What software can run Bayesian hot spot analysis?
Common choices include R packages CARBayes, INLA (via R-INLA), and brms/Stan for full Bayes; SaTScan offers a Bayesian spatial scan option. GeoBUGS (now largely replaced by Stan or INLA) was historically widely used.
What is the BYM model?
The Besag-York-Mollie (BYM) model decomposes area-level log relative risk into a spatially structured component (modeled via a conditional autoregressive prior) and an unstructured overdispersion term. This separation allows the model to detect both localized spatial clusters and global heterogeneity.
How do I choose the risk threshold c and decision threshold tau?
The choice is context-dependent. In public health, RR > 1.5 with P > 0.80 is common for alerting; for stricter surveillance, RR > 2 with P > 0.95 may be preferred. Sensitivity analyses across multiple threshold combinations are recommended to assess robustness.
Can Bayesian hot spot analysis handle covariates?
Yes. Covariates (e.g., deprivation index, age structure) can be included in the linear predictor of the hierarchical model. Residual spatial clustering after covariate adjustment is then examined, isolating unexplained hot spots from known risk-factor patterns.
Sources
- Lawson, A. B. (2018). Bayesian Disease Mapping: Hierarchical Modeling in Spatial Epidemiology (3rd ed.). CRC Press. ISBN: 978-1138575424
- Clayton, D., & Kaldor, J. (1987). Empirical Bayes estimates of age-standardized relative risks for use in disease mapping. Biometrics, 43(3), 671-681. DOI: 10.2307/2532003 ↗
How to cite this page
ScholarGate. (2026, June 3). Bayesian Hot Spot Analysis. ScholarGate. https://scholargate.app/en/spatial-analysis/bayesian-hot-spot-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bayesian Local Indicators of Spatial AssociationSpatial analysis↔ compare
- Bayesian Spatial AutocorrelationSpatial analysis↔ compare
- Hot Spot AnalysisSpatial analysis↔ compare
- Local Getis-Ord Gi*Spatial analysis↔ compare
- Local Spatial AutocorrelationSpatial analysis↔ compare