I usually start scripting my analyses with other algorithms, such as meanfield, make sure the script is running smoothly and without errors, and only then replace meanfield with "sampling" and run the analysis (which could take then days to compute).
Currently, setting the chains parameter if algorithm="meanfield" returns an error (Error: passing unknown arguments: chains.), leading to adding and removing of all the chains parameters within a script everytime we switch between two algorithms, which can become tedious in the case of many models.
I would be nice if the chains argument would just be ignored if not relevant (there could be a message: "chains parameter ignored"). What are your opinions on this?
I usually start scripting my analyses with other algorithms, such as meanfield, make sure the script is running smoothly and without errors, and only then replace meanfield with "sampling" and run the analysis (which could take then days to compute).
Currently, setting the chains parameter if
algorithm="meanfield"
returns an error (Error: passing unknown arguments: chains.
), leading to adding and removing of all the chains parameters within a script everytime we switch between two algorithms, which can become tedious in the case of many models.I would be nice if the chains argument would just be ignored if not relevant (there could be a message: "chains parameter ignored"). What are your opinions on this?