Andrea-Havron-NOAA
commented
4 months ago - [ ] fix osa cdf calls to
+=in density functions - [ ] define
keepin def.hpp - [ ] and
keepto density_components_base and model.hpp - [ ] add and expose
get_data_vector()to rcpp_interface - [ ] add data_vector to information
- [ ] add DATA_VECTOR and DATA_INDICATOR_VECTOR tp
FIMS.cpp - [ ] all links to keep in information, model, density_components_base, and FIMS.cpp
- [ ] add osa tests
Is your feature request related to a problem? Please describe.
OSA residuals are considered best practices for stock assessments but are unavailable in FIMS at the moment. They are needed to appropriate diagnose model misfit and give clues as to how to modify the model. Pearson residuals are a common approach but have some meaningful disadvantages.
Describe the solution you would like.
Add functionality for the user to request OSA residuals from the R side.
Describe alternatives you have considered
Pearson residuals can be done but are insufficient.
Statistical validity, if applicable
Thygesen et al. introduced their use in TMB models https://doi.org/10.1007/s10651-017-0372-4
and Trijoulet et al. applied them to compositions in stock assessments https://doi.org/10.1016/j.fishres.2022.106487
Describe if this is needed for a management application
No response
Additional context
WHAM and SAM both have these integrated in. For WHAM see R code to prepare inputs here run it here, data inputs here, and usage for compositions here.