Open dinga92 opened 4 years ago
To represent advanced relationships between confounds (e.g. with named access for covariate values, and as well as named variables in formulae), it may be easier to base the generators on pyradigm. This is not urgent though.
Some desirable properties for these generators include ability to specify multiple variables, levels of confounding per variable, and formulaic way to specify multivariate confounding (like interaction terms etc)
All of them doesn’t need to happen right away, by the design must allow for adding them over time.