0.0.12rc1

This release candidate continues to optimize certain elements of NiMARE processing for its NeuroLibre preprint.

What's Changed

🛠 Breaking Changes

• Replace multiprocessing with joblib for parallelization and change n_cores default to 1 by @tsalo in #597

🐛 Bug Fixes

• Retain updated Estimator in Corrector-generated MetaResults by @tsalo in #633

Other Changes

• Reduce SCALE memory usage by @tsalo in #632
• Improve memory management in MKDAChi2 Estimator by @tsalo in #638

Full Changelog: 0.0.11...0.0.12rc1

0.0.11

This release continues fixes and improvements we've made for the NiMARE manuscript. Additionally, we are starting to dramatically refactor how NiMARE stores meta-analytic data, with the end goals of (1) synchronizing NiMARE data storage with the NIMADS standard, (2) internally representing data and results in a manner that is compatible with NeuroStore, and (3) implementing a meta-analysis model specification that can be employed with both NiMARE and Neurosynth 2.0.

What's Changed

🛠 Breaking Changes

• Replace Nilearn templates with ones stored in resources folder and unset maximum Nilearn version by @tsalo in #621
• Implement cluster mass-based cluster-level Monte Carlo correction for CBMA algorithms by @tsalo in #609
• Use scikit-learn for LDAModel by @tsalo in #607

🎉 Exciting New Features

• New diagnostics module, with post-meta-analysis Jackknife method by @tsalo in #592

👎 Deprecations

• Flag peaks2maps for removal in 0.0.13/0.1.0 by @tsalo in #616

🐛 Bug Fixes

• Only download group maps when creating dataset and raise error if no images are found for a contrast by @jdkent in #580
• Force maskers to be array images instead of proxy images by @tsalo in #588

Other Changes

• Add test steps and explicit support for Python 3.9 by @JulioAPeraza in #578
• Add major classes/functions to parent namespaces by @tsalo in #600
• Make non-user-facing utility functions semi-private and improve docs by @tsalo in #604
• Move files used by examples from tests to resources by @tsalo in #605
• Improve documentation of CBMA and add methods pages by @tsalo in #610
• Use tmpdir for memmap files instead of the NiMARE data directory by @tsalo in #599

Full Changelog: 0.0.10...0.0.11

0.0.11rc1

Release Notes

This release candidate includes certain enhancements and fixes necessary for the NiMARE software paper (https://github.com/NBCLab/nimare-paper).

Changes

• [REF] Use new function to run LDA commands (#587) @tsalo
• [FIX] Force maskers to be array images instead of proxy images (#588) @tsalo
• [FIX] only download group maps when creating dataset and raise error if no images are found for a contrast (#580) @jdkent
• [ENH] Add test steps and explicit support for Python 3.9 (#578) @JulioAPeraza

0.0.10

Release Notes

The 0.0.10 release includes a number of bug fixes and improvements. The two biggest changes are (1) support for fetching and using NeuroQuery's database, and (2) a reorganization of how Neurosynth's database is fetched and converted. Both databases now use a shared format, which separates coordinates, metadata, and features into different files with a semi-standardized naming structure. The fetching and conversion functions also now support accessing multiple vocabularies provided by these databases.

🔧 Breaking changes

• The fetching functions for the Neurosynth and NeuroQuery databases now have a data_dir parameter, instead of a path parameter.
• The Neurosynth and NeuroQuery databases are now in a very different format, so the fetching and conversion functions operate quite differently.
• The low_memory parameter employed in many NiMARE classes and functions has been replaced with memory_limit.

✨ Enhancements

• Add new methods to fetch and convert the NeuroQuery database.
• Support the new format for the Neurosynth database.
• A new decoding method for regions of interest taken from Neurosynth: ROIAssociationDecoder.
• Custom __repr__ methods in many NiMARE classes, so now printing the object will show relevant information.
• Reduce memory usage in CBMA Estimators.

🐛 Bug fixes

• Extract relevant metadata in kernel transformers for Dataset-based transform calls.
• Reference selected features instead of initial features in CorrelationDecoder.
• Separate IJK calculation from coordinate space conversion.

Changes since last stable release

• [TST] Test minimum versions of dependencies (#567) @tsalo
• [FIX] Add pytest to docs dependencies (#572) @tsalo
• [ENH, DOC] Document and use data directory approach in fetching functions (#570) @tsalo
• [TST] Reorganize tests for speed (#571) @tsalo
• [ENH] Add NeuroQuery 6308 vocab to resources (#568) @tsalo
• [DOC] Improve documentation of decoders (#506) @JulioAPeraza
• [DOC] Fill out CBMA docstrings (#564) @tsalo
• [REF] Reduce memory in CBMA Estimators (#562) @tsalo
• [FIX] Drop NQ tfidf features from manifest (#561) @tsalo
• [FIX] Separate IJK calculation from coordinate space conversion (#556) @tsalo
• [FIX] Reference selected features instead of initial features in CorrelationDecoder (#560) @tsalo
• [DOC] Replace map and threshold for MACM example (#558) @tsalo
• [FIX, DOC] Use appropriate structure in Neurosynth download example (#554) @tsalo
• [DOC] Add admonition to about.rst linking to new site (#552) @tsalo
• [DOC] Update docstrings (#551) @tsalo
• [ENH] Support new format for Neurosynth and NeuroQuery data (#535) @tsalo
• [DOC] Update citation for Enge et al. (2021) (#549) @alexenge
• [FIX] Use resample=True in IBMA examples (#546) @tsalo
• [FIX] Extract relevant metadata in kernel transformers for Dataset-based transform calls (#548) @tsalo
• [DOC] Update ecosystem figure and documentation (#545) @tsalo
• [ENH] Do not apply IBMA methods to voxels with zeros or NaNs (#544) @tsalo
• [REF] Remove unused dependencies and unimplemented workflow (#541) @tsalo
• [DOC] Change napoleon settings (#540) @tsalo
• [ENH] Add ROI association decoder (#536) @tsalo
• [ENH] Add custom __repr__ methods (#538) @tsalo
• [FIX] Update CircleCI config to fix recent bug (#537) @tsalo
• [ENH] Replace low_memory with memory_limit and reduce memory bottlenecks (#520) @tsalo

0.0.10rc2

Release Notes

This second release candidate for 0.0.10 includes a major overhaul of the Neurosynth fetching and conversion functions. The Neurosynth database now follows a very different file format, in order to match NeuroQuery's convention. We also have a new function to fetch NeuroQuery, and the Neurosynth conversion functions will work with NeuroQuery data as well.

Changes

• [ENH] Support new format for Neurosynth and NeuroQuery data (#535) @tsalo
• [DOC] Update citation for Enge et al. (2021) (#549) @alexenge
• [FIX] Use resample=True in IBMA examples (#546) @tsalo
• [FIX] Extract relevant metadata in kernel transformers for Dataset-based transform calls (#548) @tsalo
• [DOC] Update ecosystem figure and documentation (#545) @tsalo
• [ENH] Do not apply IBMA methods to voxels with zeros or NaNs (#544) @tsalo
• [REF] Remove unused dependencies and unimplemented workflow (#541) @tsalo
• [DOC] Change napoleon settings (#540) @tsalo
• [ENH] Add ROI association decoder (#536) @tsalo
• [ENH] Add custom __repr__ methods (#538) @tsalo
• [FIX] Update CircleCI config to fix recent bug (#537) @tsalo
• [ENH] Replace low_memory with memory_limit and reduce memory bottlenecks (#520) @tsalo

0.0.10rc1

Release Notes

This release prepares for NeuroHackademy. The most relevant changes are the new ROIAssociationDecoder and the custom __repr__ methods.

Changes

• [DOC] Change napoleon settings (#540) @tsalo
• [ENH] Add ROI association decoder (#536) @tsalo
• [ENH] Add custom __repr__ methods (#538) @tsalo
• [FIX] Update CircleCI config to fix recent bug (#537) @tsalo
• [ENH] Replace low_memory with memory_limit and reduce memory bottlenecks (#520) @tsalo

0.0.9

Release Notes

This release primarily improves testing and documentation, but there are a few new features as well.
The new features include (1) a new "merge" method for Datasets, to combine two Datasets into a new one, (2) a new ImageTransformer class for generating new images from existing ones of different types, (3) a new "inspect" method for Correctors, to make it easier for users to see what kinds of multiple comparisons correction methods are available for a Corrector type and a specific MetaResult.

Special thanks to @alexenge for the new ALE subtraction and conjunction analysis example and to @JulioAPeraza for enhancing NiMARE's docstrings with "versionadded" and "versionchanged" directives!

Changes

• [TST, REF] Address pandas SettingWithCopyWarnings (#534) @tsalo
• [ENH] Accept multiple annotations files in Neurosynth converter (#531) @tsalo
• [ENH] Add Corrector.inspect class method (#530) @tsalo
• [TST] Cover cases where some studies are missing coordinates (#527) @tsalo
• [ENH] Add drop_invalid option to Dataset.get() and Estimator.fit() (#526) @tsalo
• [ENH] Support lists of targets in ImageTransformer (#518) @tsalo
• [DOC] Add versionadded and versionchanged directives to docstrings (#501) @JulioAPeraza
• [DOC] Add example for ALE subtraction and conjunction analysis (#519) @alexenge
• [ENH] Add merge method to Dataset class (#517) @tsalo
• [ENH] Add ImageTransformer class (#513) @tsalo
• [ENH] Add overwrite option to transform_images (#509) @tsalo

0.0.9rc2

Release Notes

This release candidate adds the ImageTransformer class for the OHBM tutorial.

Changes

• [ENH] Add ImageTransformer class (#513) @tsalo

0.0.9rc1

Release Notes

This release candidate adds an overwrite option in preparation for the OHBM 2021 NiMARE tutorial.

Changes

• [ENH] Add overwrite option to transform_images (#509) @tsalo

0.0.8

Release Notes

This release includes a number of bug-fixes, along with enhancements to how many tools within NiMARE implement low-memory options. In addition, we have renamed the CBMA estimators' null methods. The "analytic" method is now "approximate" and the "empirical" method is now "montecarlo".

Changes

• [REF] Rename CBMA null distribution generation methods (#494) @tsalo
• [FIX] Add informative error when NeuroVault collection is not found (#500) @tsalo
• [ENH] Support symmetric GCLDA topics with more than two subregions (#499) @tsalo
• [DOC] Add sphinx-copybutton to docs requirements (#502) @tsalo
• [ENH] Incorporate information about valid masking approaches into IBMA Estimators (#495) @tsalo
• [FIX] Deal with extreme t-values in t_to_z by truncating associated p-values (#498) @tsalo
• [TST] Add flake8-isort to test dependencies (#493) @tsalo
• [REF] Miscellaneous GCLDA cleanup (#486) @tsalo
• [DOC] Add new functions and classes to API documentation (#490) @tsalo
• [ENH] add images_to_coordinates (#446) @jdkent
• [ENH] Add check_type function (#480) @tsalo
• [REF] Add low_memory option to Estimators and add function for moving metadata from Dataset to DataFrame (#476) @tsalo
• [FIX] Set Dataset.basepath using absolute path (#474) @tsalo
• [FIX] Find common stem in find_stem instead of largest common substring (#472) @tsalo
• [FIX] Replace misspelled "log_p" with "logp" (#468) @tsalo
• [FIX] Assume non-symmetric null distribution in ALESubtraction (#464) @tsalo
• [TST] Add memmap test. (#463) @jdkent
• [REF] Write temporary files to the NiMARE data directory (#460) @tsalo
• [REF] Use saved MA maps, when available, in CBMA estimators (#462) @tsalo
• [FIX] Neurovault name collisions (#457) @jdkent
• [FIX] Update niftimasker in dataset blob (#459) @jdkent
• [FIX] Add work-around for maskers that do not accept 1D input (#455) @jdkent
• [ENH] Add low-memory option for kernel transformers (#453) @tsalo
• [ENH] add function to convert neurovault collections to a NiMARE dataset (#432) @jdkent
• [FIX] Ensure IBMA results have the expected number of dimensions (#450) @jdkent
• [STY, TST] Add flake8-docstrings to requirements (#435) @tsalo