Version 3.0alpha7 - 2/11/25

• (happy) Trying out some new approaches for aliased correlation.
• (happy) Put in some initial support for multiecho data.
• (happy) Updated argument parser, adjusted some argument names.
• (package) Various modernizations and cleanups.

Version 3.0alpha6 - 1/28/25 happy bugfix

• (happy) Fixed a major bug in motion filtering (thank you to Sukru Demiral for reporting it!)

Version 3.0alpha5 - 1/25/25

• (happy) Fixed a typo in argument description.
• (atlasaverage) Added the ability to segment a template atlas based on values of the map being summarized.
• (docs) Significantly beefed up happy section.
• (reference) Constructed a new version of the JHU LVL1 atlas where each territory is split into an early and late subterritory, based on the median delay in the HCP-YA dataset.
• (package) Fully converted to using pyproject.toml for packaging.

Version 3.0alpha4 - 1/21/25 bug fix, significance mask enhancement

• (rapidtide, tidepool) Both now only use neglog10p map to dynamically make significance masks.
• (rapidtide, Docker) rapidtide should now properly report it's version number in a container even if it is in prerelease.

Version 3.0alpha3 - 1/16/25

• (rapidtide) Refine delay is now on by default. Use --norefinedelay to disable it. --refinedelay still exists for compatibility, but does nothing.
• (rapidtide) Now outputs a map of fit significance (-log10(p)). In response to #174.
• (showxcorrx) Fixed some issues with axis labelling.

Version 3.0alpha2 - 12/30/24 Critical bug fix and an enhancement

• (happy, rapidtide) Now do (correct) checking to see if running in a container that now handles both Docker and Singularity/Apptainer properly. Thank you to Derek Monroe (https://github.com/dcmonroe) for the catch and the fix!
• (package) Now includes codespell checking thanks to Yaroslav O. Halchenko (https://github.com/yarikoptic)!

Version 3.0alpha1 - 12/20/24 First alpha of rapidtide 3.0

• (rapidtide) The --fixdelay option has been split into two options. --initialdelay DELAY lets you specify either a float that sets the starting delay for every voxel to that value, or a 3D file specifying the initial delay for each voxel. --nodelayfit determines whether the delay can be adjusted from its initial value. Closes #171. KNOWN ISSUE: If you supply an initial delay map, instead of using the global mean, you should use the delays to make your first stage regressor. Currently that is not the case.
• (rapidtide) Reorganized command line options and adjusted the default values.
• (rapidtide) Help output now shows the filter ranges.
• (rapidtide, retroglm) Added delay refinement using the ratio of the fit coefficients of the regressor and its time derivative.
• (rapidtide, retroglm) Fixed a bad shared memory leak.
• (retroglm) Significantly enhanced logging.
• (retroglm) Added canary files.
• (rapidtide) Implemented delay map patching.
• (rapidtide) Write out individual EV timecourses.
• (Docker) Cleaned up some internal variables.
• (Docker) Improved build and testing scripts.
• (io) Added function to compare nifti files with some tolerance.
• (docs) Automated more table generation.
• (package) Merged some dependabot PRs.
• (package) Fixed a fairly big, but not necessarily impactful bug. mlregress returned R2, not R, so anything referring to the R of a fit was actually squared (R was actually R2, R2 was actually R4).

Version 2.9.9.5 - 11/15/24 Deployment fix

• (deployment) New idea - split the README.rst file to remove potentially offending reference stuff.

Version 2.9.9.4 - 11/15/24 Deployment fix attempt

• (deployment) Attempt to fix deployment by essentially giving up.

Version 2.9.9.3 - 11/15/24 deployment bump

• (deployment) Yet another attempt to fix deployment.