From September 26-30, 2022, the DECIDE project partners at the University of Copenhagen (UCPH) hosted their first annual DECIDE workshop on state space models for animal health surveillance. The modeling workshops will take place annually and are open for members of the DECIDE consortium.
For this first workshop, 16 members of the DECIDE consortium, including workshop instructors Anders Ringgaard Kristensen, Dan Børge Jensen (UCPH, DECIDE) and Leonardo Victor de Knegt (UCPH), gathered at the University of Copenhagen, Denmark, to dedicate 5 days to Gaussian state space models (dynamic linear models). These workshops are part of DECIDE’s WP2 (Models), where the models and algorithms, i.e. the backbone of the decision support tools, are developed and validated.
As part of the DECIDE project, participants were expected to bring their own time series data to the workshop and get started with project related modeling. After lectures and small in-plenum exercies in the mornings, participants spent their afternoons by working on their own DECIDE data, supervised by the instructors.
The topics covered in this first workshop were:
- Linear algebra, brush up
- DLM, basic concepts
- First order univariate DLMs
- Estimation of variance components
- Univariate DLM with linear trend component
- Univariate DLM with Spline function
- Modeling seasonal or diurnal patterns: Harmonic waves
- Introduction to multivariate DLMs
- Early warning systems based on DLMs
Upcoming annual workshop topics:
Multi-level models
Generalized dynamic linear models (non-Gaussian)
Early detection by use of state space models
While the workshops are crucial for the joint work and progress within the project, they are also great networking spaces – strengthening collaboration, knowledge transfer, and personal connections within the DECIDE consortium.
We are looking forward to next year’s workshop. Thank you to UCPH for planning and hosting this great event!
See some impressions of the workshop below: