Relationship between Dataset and MetaAnalysis? about neurostore

[tyarkoni]

The MetaAnalysis database currently does not appear to reference the Dataset database, but in my mind I think MetaAnalysis should have an explicit relationship with the Dataset database, since the meta-analysis will perform operations on a dataset object.

Yes I think that's right. I don't think the current setup is by design; I just don't think I'd thought far enough ahead.

Should data in MetaAnalysis have the same content as nimads_data in Dataset?

Probably this should just be a foreign key to the dataset table (a Dataset can be used by many analyses, but an analysis can only have one input Dataset).

We also need a JSON field to store the analysis specification (e.g., parameters that get passed to NiMARE). Maybe that can be called config or spec or something.

If nimads_data already contains the studies/analyses/points/images necessary for the meta-analysis, does this section contain redundant information?

Unclear... definitely we need to store weights somewhere, and those wouldn't go in the Dataset. But arguably they should go in the JSON analysis specification rather than having a relational representation in the DB. We've already broken the link between the Dataset and the other objects, so it would be weird to reintroduce it here. One static JSON file containing everything needed to configure the analysis to work with the Dataset is probably the way to go.

[jdkent]

closing for cleanup