Tal Yarkoni

The intended usage pattern for `fit()` is actually to always pass in a `Dataset` (that will do the reshaping for you automatically). The only reason `**kwargs` are accepted, and I'm...

Oh, but note also that there's an `ensure_2d` helper in `stats` you can call if you want to make sure your variables are in the right shape to pass to...

> Are Datasets going to still be the preferred inputs for vectorized calls (e.g., neurostuff/NiMARE#211)? It's probably worth doing a bit of profiling to see how much overhead the `Dataset`...

I was thinking about this some more, and an alternative approach that might be more robust and flexible is to add an API route that returns different versions for a...

@adelavega you may want to try running with a hard limit of 2GB of memory to see what happens. If it takes longer to fit each model, but estimation still...

Agreed—the one potential hitch (though it may not be an issue in practice) is that the pretrained models require an input of fixed size, so we currently apply the `ImageResizingFilter`...

Re: `Replace`, I agree that this would be a good one, but for it to be useful we'd need to have a way of showing users what string values any...

WRT punctuation, I assume we're talking about reading-based designs, as punctuation doesn't really have any meaning in an auditory context, and I don't see a point in trying to encode...

This will probably require extending the NIMADS spec (either in the spec itself, or outside of it) to accommodate representation of single-subject maps and associated covariates. I.e., we don't just...