H. Tips, model tuning, caveats, and trouble shooting

Under construction

Questions to ask: Does the model fit my data? (Check the plots)

Did the model converge? (Check the gradients.)

Why doesn't my production model fit the data? (Check the starting values/modifiers for K and r.)

Do I have any parameters that are time-varying and therefore a vector? (Make sure the vector length equals SimYear.)

Error messages often pop up in the harvest control rule if the model is improperly specified. (SpBio<BMSY) NAs are not something or other. This is usually because a parameter is incorrectly specified (e.g. t0n is 2, so there are age classes with length <0 or a parameter specified as a vector is too short).

Time for simulation can be long when executing age-structured assessments. (Example times)

Error in plot comparisons when underlying operating model is different.

Using GeMS for management strategy evaluation (full-feedback), projection modeling (no-feed back management strategy evaluation), or data simulation (e.g. for fitting spawner-recruit models).

Why would one ever want to do full-feed back? The management systems and check or exacerbate biases in management controls and it is difficult to know what will happen without the feedback.

Future additions:

• Different selectivity forms.
• Economics.
• User inputted harvest control rule.
• Tagging/Growth data.
• Others noted in issue tracker.