search

neurostuff / NiMARE

Coordinate- and image-based meta-analysis in Python
https://nimare.readthedocs.io
MIT License
179 stars 58 forks source link

Support image-based meta-regression #705

Open tsalo opened 2 years ago

tsalo commented 2 years ago

Summary

PyMARE already supports meta-regression for many of its Estimators. In order to support meta-regressions in NiMARE, we basically just need to support design matrices in some of the IBMA estimators.

Additional details

This would dramatically improve the utility of the IBMA Estimators.

Next steps

  1. Decide how we want to support design matrices. Do we want X (array) + X_names (list) or a pandas DataFrame?
  2. Decide how we want to label outputs. How should we incorporate the names into the dictionary keys?
  3. Implement and test.
tsalo commented 2 years ago

I don't think we can support arrays, since we need to match variables to IDs in the Dataset. That leaves DataFrames or something more complicated.

We may also want to use Annotation objects for our design matrices- or at least have Transformers that generate design matrices from Annotations. That way, we'll be able to work well with NeuroStore, which stores anything users might want to use in a meta-regression as Annotations.