0.13.1 🌈

Changes

🧰 Maintenance

• improve syntax usage (#1005) @Zethson
• fix fknni extra (#1003) @Zethson

0.13.0 🌈

🚀 Features

• Transitioning from AnnData to EHRData
  EHRData replaces AnnData as ehrapy's core data structure to better support time-series electronic health record data.
  The key enhancement is native support for 3D tensors (observations × variables × timesteps) alongside the existing 2D matrices, enabling efficient storage of longitudinal patient data.
  A new .tem DataFrame provides time-point annotations, complementing the existing .obs and .var annotations for comprehensive temporal data description.
  While EHRData maintains full backward compatibility with AnnData's API, users can now seamlessly work with time-series data and leverage specialized methods for temporal analysis.
  Existing code using AnnData objects will continue to work, but migration to EHRData is strongly recommended to access enhanced time-series functionality.
• The preferred central data object is now EHRData (#908) @eroell
• The layers argument is now available for all functions operating on X or layers (#908) @eroell
• Update expected behaviour of io.read_fhir (#922) @eroell
• Move mimic_2, mimic_2_preprocessed, diabetes_130_raw, diabetes_130_fairlearn to ehrdata.dt (#908)
• Deprecate all ep.dt.*, refer to datasets in ehrdata (#908) @eroell
• Support Python 3.14 (#996) @Zethson
• Move kaplan_meier & cox_ph plots to holoviews (#995) @Zethson
• Longitudinal normalization (#958) @agerardy
• Add interactive ols plot (#992) @Zethson
• Longitudinal and new qc_metrics (#967) @sueoglu
• Simple Impute for timeseries (#975) @eroell
• Simple implementation of balanced sampling (#937) @sueoglu
• Add Sankey diagram visualization functions (#989) @sueoglu
• Add ep.pl.timeseries() to visualize variables over time (#994) @sueoglu
• Add GPU CI & skeleton (#998) @Zethson
• Add FAMD (#976) @Zethson
• 3D enabled implementation of ep.pp.filter_observations, ep.pp.filter_features (#953) @sueoglu
• Add time series distances (#954) @Zethson

🐛 Bug Fixes

• All green if GPU skipped (#1000) @Zethson
• Fix neighbors with timeseries (#973) @eroell
• Fix use_rep when X none (#969) @eroell
• Fix missing_values_barplot errors (#963) @sueoglu
• Fix CR notebook (#939) @Zethson

🧰 Maintenance

• Update actions (#977) @Zethson
• Cleanup simple_impute tests (#974) @eroell
• Move to ehrdata 0.0.10 (#971) @eroell
• Improved notebook CI (#959) @Zethson
• Switch to template (#960) @Zethson
• Tests for more plots (#919) @sueoglu
• Lowerbound cvxpy (#935) @Zethson
• Optimize var_metrics (#927) @Zethson
• Refactor Dask usage pattern (#926) @Zethson
• Add cover to README & remove some tokens (#923) @Zethson
• Update test coverage reporting (#918) @eroell
• Fix changelog links (#915) @Zethson
• Fixed structure of Returns in _rank_features_groups.py documentation (#911) @agerardy
• Add EHRData transition code (#897) @Zethson @eroell
• Make test that downloads dermatology dataset more robust (#906) @Zethson
• Update image source in README.md (#986) @eroell
• Fix plot docs formatting (#952) @Zethson
• Typo in the documentation of ehrapy.data.mimic_2_preprocessed (#917) @sueoglu

0.12.1 🌈

Changes

🚀 Features

• Make dowhy optional & remove medcat #903 @Zethson
• Add about page & improve citations #902 @Zethson
• Overhaul doc structure #895 @Zethson
• Move to biome & improve CI & reenable CR #890 @Zethson
• Clean up Round - cut down anndata extension functionality #880 @eroell

0.12.0 🌈

Changes

🚀 Features

• Improved KM plot data depth and functionality (#853) @aGuyLearning
• New Feature: Forestplot for CoxPH model (#838) @aGuyLearning
• Datatype Support in Quality Control and Impute (#865) @aGuyLearning
• Revamp survival analysis interface (#842) @aGuyLearning
• Improve submodule documentation (#859) @Zethson
• Update Kaplan Meier plots in survival analysis notebook (#864) @aGuyLearning

🐛 Bug Fixes

• Pass all non-nan features along desired var_names to impute (KNN) (#867) @nicolassidoux
• Remove Syntax warnings (#869) @Zethson
• Fix test_norm_power_group (#862) @Zethson

🧰 Maintenance

• Fix a typo in pl.paga_compare: pos -> pos, (#846) @VladimirShitov

0.11.0 🌈

Changes

✨ Features

• Add array type handling for normalization (#835) @eroell @Zethson

🐛 Bug Fixes

• Fix scipy array support (#844) @Zethson
• Fix casting to float when assigning numeric values; fixes normalization of integer arrays (#837) @eroell

0.10.0 🌈

🚀 Features

• Make all imputation methods consistent in regard to encoding requirements (#827) @nicolassidoux
• Add approximate KNN backend (#791) @nicolassidoux
• Improve survival analysis interface (#825) @aGuyLearning
• Python 3.12 support (#794) @Lilly-May
• Python 3.10+ & use uv for docs & fix RTD & support numpy 2 (#830) @Zethson

🐛 Bug Fixes

• move_to_x: Fix name of non-implemented argument "copy" to "copy_x", implement & test (#832) @eroell
• Contributing typo fix (#821) @aGuyLearning
• Fix miceforest (#800) @Zethson
• style: == to is for type comparison (#774) @eroell

📦 Dependencies

• loosen scipy restriction (#773) @eroell

0.8.0 🌈

Changes

🚀 Features

• remove pyyaml & explicit scikit-learn (#729) @Zethson
• Remove fancyimpute (#728) @Zethson
• Unify feature type detection (#724) @Lilly-May
• catplot (#721) @eroell
• Simplify ehrapy (#719) @Zethson
• Use all (#715) @Zethson
• Add bias detection to preprocessing (#690) @Lilly-May
• Use lamin logger (#707) @Zethson
• Add faiss backend for KNN imputation (#704) @Zethson
• Build RTD docs with uv (#700) @Zethson
• Refactor feature importance ranking (#698) @Zethson
• Simplify CI (#694) @Zethson
• Refactor outliers and IQR (#692) @Zethson
• Calculation of feature importances in a supervised setting (#677) @Lilly-May
• Speed up winsorize (#681) @Zethson
• Remove notebook prefix in tutorial URLs (#679) @Zethson
• Add cohort tracking notebook (#678) @Zethson
• Switch to uv (#674) @Zethson
• Style: typing of _scale_func_group (#727) @eroell
• Improved support of encoded features in detect_bias (#725) @Lilly-May
• Enable Synchronous dataloader write (#722) @wxicu
• Feature scaling on training set when computing feature importances (#716) @Lilly-May
• add batch-wise normalization argument (#711) @eroell
• add functools.wraps to type check (#705) @eroell
• add bias notebook to list of notebooks (#696) @eroell
• basic sampling (#686) @eroell
• add options for subitles in legend of cohorttrackers barplot (#688) @eroell
• doc fix imputation: 70 instead of 30 (#683) @eroell

🐛 Bug Fixes

• Encoded dtype to float32 instead of np.number (#714) @Zethson
• Fix feature importance warnings (#708) @Zethson
• Remove notebook prefix in tutorial URLs (#679) @Zethson
• fix name of log_rogistic_aft to log_logistic_aft (#676) @eroell

🧰 Maintenance

• Remove notebook prefix in tutorial URLs (#679) @Zethson
• Add cohort tracking notebook (#678) @Zethson
• knni amendments (#706) @eroell

0.7.0 🌈

Changes

🚀 Features

• Cohort Tracker (#658) @eroell
• change diabetes-130 datasets which are provided (#672) @eroell
• More sa functions (#664) @fatisati
• Coxphfitter (#643) @fatisati
• Implement little's test (#667) @Zethson
• Improve test design (#651) @Zethson
• Improve QC docstring (#639) @Zethson
• Refactor _missing_values calculation (#638) @Zethson

🐛 Bug Fixes

• Fix one-hot encoding tests (#644) @Zethson

0.6.0 🌈

Changes

🚀 Features

Breaking changes

• Move information on numerical/non_numerical/encoded_non_numerical from .uns to .var (#630) @eroell

Make older AnnData objects compatible using

def move_type_info_from_uns_to_var(adata, copy=False):
    """
    Move type information from adata.uns to adata.var['ehrapy_column_type'].
    
    The latter is the current, updated flavor used by ehrapy.
    """
    if copy:
        adata = adata.copy()
        
    adata.var['ehrapy_column_type'] = 'unknown'

    if 'numerical_columns' in adata.uns.keys():
        for key in adata.uns['numerical_columns']:
            adata.var.loc[key, 'ehrapy_column_type'] = 'numeric'
    if 'non_numerical_columns' in adata.uns.keys():
        for key in adata.uns['non_numerical_columns']:
            adata.var.loc[key, 'ehrapy_column_type'] = 'non_numeric'
    if 'encoded_non_numerical_columns' in adata.uns.keys():
        for key in adata.uns['encoded_non_numerical_columns']:
            adata.var.loc[key, 'ehrapy_column_type'] = 'non_numeric_encoded'
            
    if copy:
        return adata

New features

• Medcat refresh (#623) @eroell
• Rank features groups obs (#622) @eroell
• Add FHIR tutorial and simplify code (#626) @Zethson
• Add input checks for imputers (#625) @Zethson
• Removed unused dependencies (#615) @Zethson
• Refactor encoding (#588) @Zethson

🐛 Bug Fixes

• Use fixtures for preprocessing tests (#577) @Zethson

🧰 Maintenance

• Refactoring (#627) @Zethson
• Add FHIR tutorial and simplify code (#626) @Zethson
• pre-commit (#587) @Zethson
• Small edits (#599) @eroell

0.5.0 🌈

Changes

🚀 Features

• Add g-tests for rank features group (#546) @VladimirShitov
• Causal Inference with dowhy (#502) @timtreis
• Remove MuData support (#545) @Zethson

🐛 Bug Fixes

• Fixed reading format warnings (#569) @namsaraeva
• Fixed inability to normalize AnnData that does not require encoding (#568) @namsaraeva
• Fixed adata.uns["non_numericlal_columns"] being empty in mimic_2 dataset (#567) @namsaraeva