HSSM v0.2.4

What's Changed

• refactor: use fstring in error message by @cpaniaguam in #506
• Fix Param.update and add test case by @cpaniaguam in #505
• Fix validation and remove unnecessary assert statements by @cpaniaguam in #507
• Fix categorical variables bug by @AlexanderFengler in #513
• Fix compatibility with Bambi by @digicosmos86 in #516
• 537 initial vi support and tutorial by @AlexanderFengler in #545
• Fix initval setters leading to NotImplementedError by @AlexanderFengler in #547
• Workshop tutorial by @AlexanderFengler in #488
• Refactor test functions with parametrization by @cpaniaguam in #554
• Minor bug fixes by @digicosmos86 in #567
• Fix float32 vs float64 issue by @digicosmos86 in #568
• Release v0.2.4 by @digicosmos86 in #575
• Added HSSM.restore_traces() method by @digicosmos86 in #572
• Fixed error of parent parameter specification by @digicosmos86 in #576
• Revert changes to analytical likelihood function by @digicosmos86 in #583
• add working log likelihood computation to hssm class by @AlexanderFengler in #577

New Contributors

• @cpaniaguam made their first contribution in #506

Full Changelog: v0.2.3...v0.2.4

HSSM v0.2.3

What's Changed

• put in prep paper up front in readme by @AlexanderFengler in #467
• fix bug where resampling doesnt update the traces argument by @AlexanderFengler in #478
• Reconciliate re-sampling behavior by @digicosmos86 in #485
• Fix compatibility with PyMC>=5.16 by @digicosmos86 in #484

Full Changelog: v0.2.2...v0.2.3

HSSM v0.2.2

What's Changed

• Update documentation related to JAX by @digicosmos86 in #423
• Update pyproject.toml by @AlexanderFengler in #434
• Update documentation after hssm is available on conda-forge by @digicosmos86 in #428
• Allow users to specify the number of choices by @digicosmos86 in #430
• One-command support for JAX GPU by @digicosmos86 in #441
• 414 long running time of sample posterior predictive and eventual death by oom by @AlexanderFengler in #436
• Update reference to jax.config by @digicosmos86 in #443
• Fix ensure positive ndt by @digicosmos86 in #452
• fix key error bug when going through initialization, make safe priors… by @AlexanderFengler in #446
• Ready for v0.2.2 release by @digicosmos86 in #461

Full Changelog: v0.2.1...v0.2.2

HSSM v0.2.1

What's Changed

• Fixed bounds for "ddm_sdv" models by @digicosmos86 in #366
• Support for Go-Nogo and deadlines with missing data by @digicosmos86 in #358
• Improve documentation for Go-Nogo, bug fixes and test improvement by @digicosmos86 in #375
• 373 shift prior means for log logit settings by @AlexanderFengler in #377
• release pymc version to newest and adjust git action accordingly by @AlexanderFengler in #382
• Moved fixed effects before random effects in formula specifications in HSSM documentation by @digicosmos86 in #391
• Drop black in favor of ruff formatter by @digicosmos86 in #420
• change order between sz and sv in list_params by @AlexanderFengler in #408

Full Changelog: v0.2.0...v0.2.1

v0.2.0

What's Changed

This release bumps HSSM 0.2.0b1 to 0.2.0. Changes in HSSM 0.2.0 are:

The main purpose of this update is to create better defaults for priors, to reign in convergence issues that result from bad chain initializations.

Multiple prior strategies are now available, and moreover it is now possible to choose default links on glms so that parameter spaces are a priori constraint to be within the the allowable ranges (this is specifically relevant for likelihoods of the kind 'approx_differentiable').

Secondarily, plotting capabilities were added, to allow for posterior predictive plots and quantile plots (these plots are widely used for the analysis of reaction time and choice data). Plotting is still in experimental stage.

• Update ssm-simulators version by @digicosmos86 in #278
• Some cosmetic changes to the documentation by @digicosmos86 in #281
• Fix bug in extra fields by @digicosmos86 in #286
• Bumped PyMC version and limited numpy version by @digicosmos86 in #283
• Bumped ssms version so tests pass by @digicosmos86 in #296
• Posterior predictive plotting by @digicosmos86 in #303
• Create convenience plotting functionalities by @digicosmos86 in #300
• First attempt at Quantile probability plot by @digicosmos86 in #319
• Override defaults for link functions by @digicosmos86 in #320
• Added tests and documentation for quantile probability plots by @digicosmos86 in #339
• Added the ability to use safe priors for hierarchical models by @digicosmos86 in #331
• Replace negative non-decision time in likelihood computation with a small value by @digicosmos86 in #344

Full Changelog: v0.1.5...v0.2.0

v0.2.0b1

The main purpose of this update is to create better defaults for priors, to reign in convergence issues that result from bad chain initializations.

Multiple prior strategies are now available, and moreover it is now possible to choose default links on glms so that parameter spaces are a priori constraint to be within the the allowable ranges (this is specifically relevant for likelihoods of the kind 'approx_differentiable').

Secondarily, plotting capabilities were added, to allow for posterior predictive plots and quantile plots (these plots are widely used for the analysis of reaction time and choice data). Plotting is still in experimental stage.

HSSM v0.1.5

We fixed the errors in v0.1.4. Sorry for the convenience! If you have accidentally
downloaded v0.1.4, please make sure that you update hssm to the current version.

• We made Cython dependencies of this package available via pypi. We have also built
  wheels for (almost) all platforms so there is no need to build these Cython
  dependencies.

HSSM v0.1.4

What's Changed

• Added a hssm.Prior class to better handle bounds by @digicosmos86 in #245
• Major refactor of hssm.HSSM via Config by @digicosmos86 in #246
• Add issue templates by @digicosmos86 in #257
• 238 refactor organize tests by @digicosmos86 in #258
• Refactor: the Param class by @digicosmos86 in #251
• Added blackbox likelihoods for ddm and ddm_sdv models by @digicosmos86 in #267
• Fixed a bug in sample_posterior_predictive() by @digicosmos86 in #269
• Support for arbitrary additional fields by @digicosmos86 in #270
• Updated documentation for v0.1.4 by @digicosmos86 in #273

Full Changelog: v0.1.3...v0.1.4

HSSM v0.1.3

What's Changed

• Fix nuts_numpyro sampling error with regressions by @digicosmos86 in #227
• Updated examples to use normal instead of uniform distributions by @digicosmos86 in #231
• Better defaults for model.sample() by @digicosmos86 in #232
• Updated the way model configs are handled by @digicosmos86 in #234
• Documentation reorganization and various bug fixes by @digicosmos86 in #240

Full Changelog: v0.1.2...v0.1.3

HSSM v0.1.2

• Improved numerical stability of analytical likelihoods. #210 #182

• Added the ability to handle lapse distributions. #215

• Added the ability to perform prior predictive sampling. #202

• Improved model information output - now default priors provided by bambi is also printed. #211

• Added a hierarchical switch which turns all parameters into hierarchical
  when participant_id is provided in data. #197

• Parameters are now named more consistently (group-specific terms are now aliased correctly). #211

• Fixed a bug where information about which parameter is regression is incorrectly passed. #219

• Added links to Colab to try out hssm in Google Colab. #194 #199