Taylor Salo

Multiple comparisons correction with PyMARE

2

Multiple comparisons correction and permutation-based parameter estimation for the new IBMA estimators (built on top of PyMARE) are not implemented.

Cluster-corrected logp and z-statistic maps do not completely match

3

## Summary This mismatch was reported by a user via email. The issue seems to result from the fact that, in a one-tailed p-to-z transform, p values greater than 0.5...

Add `nimare_version` attribute to all classes

5

## Summary NiMARE classes (especially `Datasets`) are going to be in flux a bit until we've moved onto stable versions. This has caused problems in the past, so it would...

Synchronize NiMARE json format with NIMADS specification

2

## Summary It looks like [NIMADS](http://neurostuff.org/nimads/) ([example](https://neurostore.org/api/studies/87UtKVNzN8cA?nested=true)) has evolved and stabilized significantly, so it would be good to update how NiMARE stores and loads its JSON files. This won't affect...

Define parameters in a central location and reference those definitions in docstrings

1

## Summary Rather than redefine the same parameter/return value descriptions many times across the library, which will increase the chances of discrepancies across those descriptions, we can define these parameters...

Versioneer-generated version strings are "dirty" on RTD

1

## Summary I've noticed that the version strings on the RTD site always say "dirty", indicating uncommitted changes, even for builds tied to specific releases. This isn't a big problem,...

Using pre-made MA maps in a meta-analysis is more memory-intensive and slower than making them in `fit`

6

## Summary We have an argument, `return_type`, in `KernelTransformer.transform` to determine how the modeled activation (MA) maps should be returned. The options are `array`, `image`, and `dataset`. With `dataset`, the...

Should the maximum statistic null distributions be separated by sign for two-sided tests?

2

One question I have is whether the maximum statistic null distributions should be separated by sign for two-sided tests. For example, for the MKDAChi2, which can have positive and negative...

Use nilearn masker memory parameters instead of safe_transform

2

## Summary The nilearn masker objects have `memory` and `memory_level` parameters. I don't have much experience with them, but I think they could be used as an alternative to our...

Create DatasetSearcher class for transforming Datasets

1

## Summary Per today's discussion, one thing that would make it easier to simplify the Dataset object would be to move the Dataset's searching methods into a new Transformer-like class....