Added the ability to use safe priors for hierarchical models #331
Conversation
| @@ -24,6 +24,7 @@ | |||
| import seaborn as sns | |||
| import xarray as xr | |||
| from bambi.model_components import DistributionalComponent | |||
| from bambi.transformations import transformations_namespace | |||
There was a problem hiding this comment.
what is that one for actually?
There was a problem hiding this comment.
I think there is a default namespace that has to be included. This can be found in Bambi source code here
| @@ -531,3 +654,14 @@ def _make_default_prior(bounds: tuple[float, float]) -> bmb.Prior: | |||
| return bmb.Prior("TruncatedNormal", mu=lower, lower=lower, sigma=2.0) | |||
| else: | |||
| return bmb.Prior(name="Uniform", lower=lower, upper=upper) | |||
|
|
|||
|
|
|||
| def merge_dicts(dict1: dict, dict2: dict) -> dict: | |||
There was a problem hiding this comment.
strictly speaking this is not a merge right?
seems like if key in merged but instances are not dicts, you override the value of dict1 with whatever you find in dict2.
So maybe something like override_dict ?
There was a problem hiding this comment.
This actually raises a question: should the override be recursive? For example, if 1|participant_id has a defined prior, with only sigma having a defined hyperprior, should we also add our default mu? It seems that if we don't then a bambi or pymc default will be applied
There was a problem hiding this comment.
I would say no, because we would in fact not cover the case in all generality even then.
The logic would only apply if the user defined distribution 1|participant_id matches the one for which we have hyper-priors available.
If the user defines their own prior, they should define it all the way down I would say.
pyproject.toml
Outdated
| @@ -23,14 +23,14 @@ numpy = ">=1.23.4,<1.26" | |||
| onnx = "^1.12.0" | |||
| jax = "^0.4.0" | |||
| jaxlib = "^0.4.0" | |||
| ssm-simulators = "0.5.1" | |||
| ssm-simulators = "0.5.3" | |||
There was a problem hiding this comment.
can be bumped to 0.6.1 now
No description provided.