Template Model Building (TMB) is amongst the newest modern tools for efficiently fitting highly parameterized stock assessment and other models with random effects.
TMB is a new tool which supports the same algorithms as AD Model Builder, in which most current assessment models are build. TMB is designed to be very efficient by combining modern, carefully selected, and well suported libraries (e.g,cppAD, R, Eigen, CHOLMOD).
This course will be taught at an advanced level showing how fisheries assessment models can be written and fitted in TMB. The course is intended for those who already code their own assessment models and who are involved in the provision of advice in fisheries assessment.
After the course, participants are able to:
- run the State-space Assessment Model (SAM) with their own data
- modifiy SAM and utilize the tools shared via Github to produce standard outputs for modified models
- know how to implement new models in TMB (including spatial models and age-based models conditioned on length data)
- understand the integrated-assessment-model paradigm and how such models could be coded in TMB (e.g., the incorporation of tagging experiment results in the assessment model estimation)
The course will also include a section on model selection, allowing users to evaluate the evidence for models that make different model structure assumptions related to elements such as the shapes of time-varying selectivities and natural mortality, catch underreporting, and stock-recruitment relationships.
Strategies for projecting population trends in light of time-varying model parameters will also reviewed and discussed.
