Geneva, february 13 - 17, 2006

Day 1. 13 february 2006

Opening of the course. Environment and health data management in Europe: current state and evolution. Analysis and use of such data as a basis for defining public health policy.

Introduction. Environment and health data: why geostatistics? Types of data. Data used in the course. Documentation and softwares.

Exploratory analysis of exposure data. Data import and visualization in STIS software. Summary statistics and histograms. Data transformation. Detection of spatial outliers. Exercises.

Exploratory analysis of health data. Data import and visualization in STIS software. Summary statistics and histograms. Data transformation. Small number problem and rate smoothing. Detection of spatial clusters and outliers. Exercises.

Day 2. 14 february 2006

Description of spatial patterns. Concepts of semi-variogram cloud, semi-variogram, correlogram. Analysis of direction-dependant variability. Semi-variographic map. Exercises.

Modeling the spatial variability. Deterministic vs probabilistic modeling. Concept of random function. Modeling the semi-variogram. Anisotropy modeling. Exercises.

Spatial prediction. Concept of kriging. Change of support. Cross-validation and jackknife. Aggregation and ecological regression. Exercises.

Day 3. 15 february 2006

Geostatistical smoothers and scale-dependent correlations. Binomial cokriging and Poisson kriging. Nested semivariogram models. Concept of scale-dependent correlation. Application to smoothing of cancer mortality rates. Detection of health disparities. Exercises.

Accounting for secondary information in kriging. Exhaustive secondary information: kriging within strata, simple kriging with varying local means, kriging with an external drift, colocated kriging. Better sampled secondary information: cokriging, cross semivariogram estimation and modeling. Exercises.

Day 4. 16 february 2006

Risk mapping and decision-making. Parametric vs non parametric approaches. Indicator kriging. Exercises. Accounting for uncertainty in decision-making. Quality of uncertainty models. Case studies: arsenic in groundwater (Michigan, USA), airborne cadmium contamination.

Day 5. 17 february 2006

Stochastic simulations. Simulation vs estimation. Why stochastic simulations? Simulation algorithms. Generation of null spatial models (LISA test). Exercises.

Space-time geostatistics. Approaches available: 1. space-rich time-poor information, 2. space-poor time-rich information, 3. space-rich time-rich information. A space-time model. Modeling and propagation of uncertainty. Case study.