Combine efforts with NiMARE

[jdkent]

@FeHoff Would y'all be open to combining efforts to porting some of the algorithms you're using here into NiMARE?
I've got a start here: (neurostuff/NiMARE#1009)

I don't think NiMARE's trying to do everything that JALE is trying to do (no GUI for instance) and perhaps there could be a compatibility layer here to get outputs and stats that NiMARE does not compute/workflows that nimare does not create.

Those algorithms could get extra exposure through the NiMARE interface and through neurosynth compose (and we want to get the citation flow on neurosynth down so when people select algorithms they get proper citations). Also implementations of general methods could benefit algorithms beyond ALE (probabilistic ALE could be packaged to be applicable to MKDA as well; and the thresholding for ALEContrast could be applied to MKDAChi2 too). So the benefits could compound/opportunities expand if implementations are combined.

[FeHoff]

I like the idea and will bring it up in the next group meeting. Thanks

[jdkent]

under the current architecture, NiMARE makes the balanced contrast and probabilistic ALE faster, but the contrast and predictive ALE are currently slower, with more developer time, I'm sure those differences could be ameliorated. NiMARE has some machinery to be general purpose, resulting in the relative slowness, it's an open question on how much of a difference it makes with realistic permutation choices.

https://github.com/jdkent/JALE_NiMARE_comparison/blob/main/results/figures/speed_ratio_summary.png

[FeHoff]

That looks great @jdkent ! We are actually in the process of making the balanced contrasts and probabilistic ALE cite-able and will be very happy to see those distributed with NiMARE when they're published. The same is true for the ALE clustering approach we are writing up now. We'll notify you as soon as it's out. As predictive ALE is published, feel free to include it directly. Thanks!