Closed edvinf closed 10 months ago
This turns out to be a bug in Reca. I have reported it to maintainers and will try to figure out a workaround.
The flag old.version on GlobalParameters has been used in experimental Reca code, and a check for this remains in a piece of code executed if non-linear is chosen as lgamodel. It can safely be set to zero for executions of Reca, and I will implement a workaround that sets old.version on GlobalParameters before calling Reca. Future releases of Reca will remove any checks on old.version.
When the non-linear option is chosen for ParameterizeReca, one will often get the error-message 'argument is of length zero' after the process has been running for a while. This was experiences when testing the optinon for Mackerell 2022-fishery, and NEA haddock for the 2022-fishery. The option is not necessarily intended to be used for either of these stocks. It may be a poor model fit, and may not converge, but we should be able to provide an error message consistent with other model failures if that happens.