Recent development for chemical & biological risk assessment has gradually moved from classical toxicological studies relying on animal experimental data to incorporate a population health approach by taking into account multiple determinants of health and their interactions to better inform risk management decision making. New statistical risk assessment methodologies are needed to integrate multiple sources of exposure uncertainties on the pathway leading to multiple adverse health effects.
The earth has witnessed global climate change for the past few decades. More frequent, intense, and longer lasting heat waves, as well as increasing intensity of short duration precipitation events or drought are predicted in the near future. Health impacts due to extreme weather variations and their possible extents is of great importance for government health policy decision making.
In recent decades, one of most popular tools for disease surveillance is to use clustering analysis methods. Spatial clusters are groups of spatial units with similar characteristics within a group but with dissimilar objects among groups. Clustering analysis is widely used in many applications such as disease surveillance and image process, and it can also apply to categorize gene with similar functionalities.
Regarding environmental health risks, we aim to:
- Integrate multiple sources of environmental exposures and health outcomes or biomarkers to derive reference dose for regulatory purpose;
- Assess health risk impact due to climate change;
- Investigate air pollution and cardiovascular risk;
- Conduct vector-borne disease health risk assessment.
Regarding statistical methods for disease surveillance, we aim to address several important issues in clustering analysis:
- Create multiple dimensional data with the cluster detection method for big data;
- Identify clusters of arbitrary shapes with algorithms and cluster detection procedures to overcome the barriers of distance measures that tend to find circular cluster of small sizes;
- Analyze spatial noisy data without adjusting random effects that leads to poor detection result.