Suggestion for Improving Prior Setup in Hierarchical Models

[gingjehli]

It would be helpful if one could specify priors for classes such as intercepts, coefficients, and covariance matrices across all hierarchical levels of a model (e.g., groups, treatments, or subjects). This way, one can ensure that the chosen priors are applied consistently throughout the model, which makes results more reliable.

Also, it would be great to incorporate warning messages if one hasn't set priors for every parameter in a provided regression equation. A warning message could pop up to remind that some priors are missing, and the software is going to use the default ones instead. This feature would prevent from accidentally overlooking a setting and help avoid the issues that come with unintended default priors.

Adding these features would make the modeling process smoother and more user-friendly, giving better control and clarity over how our models are built.

[AlexanderFengler]

Hey @gingjehli ,

we will look into this. Adding the warnings you suggest is trivial, we can definitely do that.
Priors for parameter-groups will demand a bit more thought, but is doable as well. Thank you for raising those points.