Problem definition
When sampling from certain models/datasets there is a difficulty accepting new samples from the proposals due to the use of the group level covariance matrix at the individual subject level. Setting epsilon can help with this, but a nice method to avoid manually running and rerunning with new epsilon values would be to automatically tune the epsilon value based on current acceptance rate.
Describe the solution you'd like
A new "auto" value for epsilon that implements the auto tuning capability described above
Additional context
A dataset provided by Scott/Aline can be a good candidate for testing
gjcooper
commented
3 years ago
Problem definition When sampling from certain models/datasets there is a difficulty accepting new samples from the proposals due to the use of the group level covariance matrix at the individual subject level. Setting epsilon can help with this, but a nice method to avoid manually running and rerunning with new epsilon values would be to automatically tune the epsilon value based on current acceptance rate.
Describe the solution you'd like A new "auto" value for epsilon that implements the auto tuning capability described above
Additional context A dataset provided by Scott/Aline can be a good candidate for testing