PyBIDS/StatsModels improvements

[tyarkoni]

PyBIDS/BIDS-StatsModels Improvements

Instructions and template for final projects.

• Tal Yarkoni, UT-Austin
• Chris Markiewicz, Stanford
• Alejandro de la Vega, UT-Austin
• Dylan Nielson, NIH

---

Brief description of what was accomplished with this project

To facilitate estimation of a wider range of models via fitlins, and enable AFNI support (see #1), changes are required to PyBIDS and the BIDS-StatsModels spec. We implemented a number of major changes/improvements, including:

• New model_design module containing an abstract specification of statistical models (with classes like Term, VarianceComponent, etc.). This will serve as an intermediary between BIDSVariableCollection instances and fitlins interfaces.
• New TransformerManager class that generalizes the existing transformation application procedure to allow registration and execution of arbitrary callables.
• Various transformation-related improvements (see ENH: Transformation-related improvements bids-standard/pybids#541) that were in the BIDS-StatsModels spec and are necessary to support mixed-effects modeling, but were previously supported in PyBIDS.
• Several revised/improved sections of the BIDS-StatsModels spec, mostly focused on the Model section.

---

Resources

PyBIDS

---