One of the hallmarks of factor analysis is that it allows for non-orthogonal latent variables.
In R for example this feature is accessible via the rotation parameter of factanal.
Is there any such provision for sklearn.decomposition.FactorAnalysis? Clearly it's not among the arguments - but maybe there is another way to achieve this?
Sadly I have been unable to find many examples of usage for this function.
Interesting question. I think there are no rotations implemented indeed - see this issue.
Maybe this implementation is what you are looking for.