Orange3-MRSD 0.1.0__tar.gz

orange3_mrsd-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Lan Umek
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

orange3_mrsd-0.1.0/MANIFEST.in

@@ -0,0 +1,8 @@
+include README.md
+include LICENSE
+recursive-include orangecontrib/subgroups/widgets/icons *.svg
+recursive-include orangecontrib/subgroups/datasets *.tab *.csv
+recursive-include orangecontrib/subgroups/mrsd *.py *.md
+recursive-include tools *.py
+recursive-include doc *.md
+recursive-include doc/widgets/images *.png

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/PKG-INFO

@@ -0,0 +1,103 @@
+Metadata-Version: 2.4
+Name: Orange3-MRSD
+Version: 0.1.0
+Summary: Orange add-on for discovering subgroups in data with several dependent variables (MRSD).
+Home-page: https://github.com/lan-umek/orange3-mrsd
+Author: Lan Umek
+License: MIT
+Keywords: orange3 add-on,subgroup discovery,multiple responses,MRSD,exceptional model mining
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Information Analysis
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: Orange3>=3.34
+Requires-Dist: numpy>=1.23
+Requires-Dist: pandas>=1.5
+Requires-Dist: scipy>=1.9
+Requires-Dist: scikit-learn>=1.2
+Requires-Dist: pyqtgraph
+Provides-Extra: excel
+Requires-Dist: openpyxl; extra == "excel"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Dynamic: author
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: keywords
+Dynamic: license
+Dynamic: license-file
+Dynamic: provides-extra
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# Orange3-MRSD
+
+An [Orange](https://orangedatamining.com) add-on for **Multiple Responses
+Subgroup Discovery (MRSD)**: finding groups of instances, described by the
+independent variables **X**, in which several dependent variables **Y** – or
+their relation to X – behave differently from the rest of the data.
+
+The add-on wraps the MRSD Python package (vendored in
+`orangecontrib/subgroups/mrsd`) and adds widgets for setting variable roles,
+running the algorithms, inspecting, combining, visualising and saving the
+subgroups.
+
+## Widgets (category *Subgroups*)
+
+| | |
+|---|---|
+| **Data** | Subgroup Datasets, Variable Roles |
+| **Algorithms** | MRSD-PT (partitioning and tables), MRSD-HC (hierarchical clustering and classification), MRSD-C2 (classification in both spaces: swapping, semi-supervised, ranking), MRSD-CCA, MRSD-EMM, MRSD-MV (multi-view), MRSD-MI, MRSD-Rules, MRSD-Ensemble |
+| **Results** | Subgroup Browser, Subgroup Consolidation, Subgroup Classifier, Subgroup Profile, Subgroup Overlap, Save Subgroups, Load Subgroups |
+
+Every algorithm widget shows the subgroups with their rule in X and prototype
+in Y, has algorithm-specific views of the selected subgroup (contingency
+table, dendrograms, classifiers, correlations …), accepts Orange learners and
+sends the selected subgroup's rows, membership data and membership model on
+to other Orange widgets.
+
+Documentation: [doc/index.md](doc/index.md) – start with
+[Getting started](doc/getting-started.md).
+
+## Installation
+
+Orange 3.34 or newer is required.
+
+```
+pip install Orange3-MRSD        # into the environment that runs Orange
+```
+
+or from a clone: `pip install -e .`.  Excel export additionally needs
+`openpyxl`.
+
+## Development
+
+```
+python -m pytest tests -q        # backend tests run without Orange; widget tests need Orange and QT_QPA_PLATFORM=offscreen
+python tools/make_docs.py        # regenerate doc/ (parameter tables come from params.py)
+python tools/make_datasets.py    # regenerate the bundled datasets
+python tools/sync_mrsd.py        # copy the MRSD package from ../subgroups into orangecontrib/subgroups/mrsd
+```
+
+* The widgets build their parameter GUI from `orangecontrib/subgroups/params.py`;
+  `tests/test_params.py` fails if a field of an algorithm's `Config` dataclass is
+  not covered or a parameter name clashes with a widget attribute.
+* Orange caches the widget descriptions of an add-on and refreshes them only
+  when `orangecontrib/subgroups/widgets/__init__.py` changes: after changing a
+  widget's inputs, outputs or metadata bump `WIDGET_REGISTRY_STAMP` there.
+
+## Status
+
+The algorithms and the non-GUI part of the add-on are covered by tests.  Most widgets have been tried interactively on Orange 3.40 (Windows); some of
+the newer views (e.g. the dendrograms of MRSD-MV, cell selection in Subgroup
+Overlap) and other Orange versions and platforms have not been checked yet.
+
+## License
+
+MIT – see [LICENSE](LICENSE).

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/SOURCES.txt

@@ -0,0 +1,182 @@
+LICENSE
+MANIFEST.in
+README.md
+setup.py
+Orange3_MRSD.egg-info/PKG-INFO
+Orange3_MRSD.egg-info/SOURCES.txt
+Orange3_MRSD.egg-info/dependency_links.txt
+Orange3_MRSD.egg-info/entry_points.txt
+Orange3_MRSD.egg-info/not-zip-safe
+Orange3_MRSD.egg-info/requires.txt
+Orange3_MRSD.egg-info/top_level.txt
+doc/concepts.md
+doc/getting-started.md
+doc/index.md
+doc/widgets/load-subgroups.md
+doc/widgets/mrsd-c2.md
+doc/widgets/mrsd-cca.md
+doc/widgets/mrsd-emm.md
+doc/widgets/mrsd-ensemble.md
+doc/widgets/mrsd-hc.md
+doc/widgets/mrsd-mi.md
+doc/widgets/mrsd-mv.md
+doc/widgets/mrsd-pt.md
+doc/widgets/mrsd-rules.md
+doc/widgets/save-subgroups.md
+doc/widgets/subgroup-browser.md
+doc/widgets/subgroup-classifier.md
+doc/widgets/subgroup-consolidation.md
+doc/widgets/subgroup-datasets.md
+doc/widgets/subgroup-overlap.md
+doc/widgets/subgroup-profile.md
+doc/widgets/variable-roles.md
+doc/widgets/images/load-subgroups.png
+doc/widgets/images/mrsd-c2-classifier.png
+doc/widgets/images/mrsd-c2-measures.png
+doc/widgets/images/mrsd-c2-prototype.png
+doc/widgets/images/mrsd-c2.png
+doc/widgets/images/mrsd-cca-canonical-space.png
+doc/widgets/images/mrsd-cca-classifier.png
+doc/widgets/images/mrsd-cca-measures.png
+doc/widgets/images/mrsd-cca-prototype.png
+doc/widgets/images/mrsd-cca.png
+doc/widgets/images/mrsd-emm-correlations.png
+doc/widgets/images/mrsd-emm-measures.png
+doc/widgets/images/mrsd-emm-prototype.png
+doc/widgets/images/mrsd-emm.png
+doc/widgets/images/mrsd-ensemble-consensus.png
+doc/widgets/images/mrsd-ensemble-measures.png
+doc/widgets/images/mrsd-ensemble-prototype.png
+doc/widgets/images/mrsd-ensemble.png
+doc/widgets/images/mrsd-hc-classifier.png
+doc/widgets/images/mrsd-hc-dendrogram.png
+doc/widgets/images/mrsd-hc-measures.png
+doc/widgets/images/mrsd-hc-prototype.png
+doc/widgets/images/mrsd-hc.png
+doc/widgets/images/mrsd-mi-measures.png
+doc/widgets/images/mrsd-mi-mutual-information.png
+doc/widgets/images/mrsd-mi-prototype.png
+doc/widgets/images/mrsd-mi.png
+doc/widgets/images/mrsd-mv-classifier.png
+doc/widgets/images/mrsd-mv-dendrograms.png
+doc/widgets/images/mrsd-mv-measures.png
+doc/widgets/images/mrsd-mv-prototype.png
+doc/widgets/images/mrsd-mv.png
+doc/widgets/images/mrsd-pt-contingency-table.png
+doc/widgets/images/mrsd-pt-measures.png
+doc/widgets/images/mrsd-pt-prototype.png
+doc/widgets/images/mrsd-pt.png
+doc/widgets/images/mrsd-rules-measures.png
+doc/widgets/images/mrsd-rules-prototype.png
+doc/widgets/images/mrsd-rules-rule.png
+doc/widgets/images/mrsd-rules.png
+doc/widgets/images/save-subgroups.png
+doc/widgets/images/subgroup-browser.png
+doc/widgets/images/subgroup-classifier.png
+doc/widgets/images/subgroup-consolidation.png
+doc/widgets/images/subgroup-datasets.png
+doc/widgets/images/subgroup-overlap.png
+doc/widgets/images/subgroup-profile.png
+doc/widgets/images/variable-roles.png
+orangecontrib/__init__.py
+orangecontrib/subgroups/__init__.py
+orangecontrib/subgroups/backend.py
+orangecontrib/subgroups/common_measures.py
+orangecontrib/subgroups/consensus.py
+orangecontrib/subgroups/export.py
+orangecontrib/subgroups/inspection.py
+orangecontrib/subgroups/params.py
+orangecontrib/subgroups/worker.py
+orangecontrib/subgroups/datasets/__init__.py
+orangecontrib/subgroups/datasets/clinical.csv
+orangecontrib/subgroups/datasets/clinical.tab
+orangecontrib/subgroups/datasets/customers.csv
+orangecontrib/subgroups/datasets/customers.tab
+orangecontrib/subgroups/datasets/planted_rule.csv
+orangecontrib/subgroups/datasets/planted_rule.tab
+orangecontrib/subgroups/datasets/planted_shift.csv
+orangecontrib/subgroups/datasets/planted_shift.tab
+orangecontrib/subgroups/datasets/pure_noise.csv
+orangecontrib/subgroups/datasets/pure_noise.tab
+orangecontrib/subgroups/datasets/students.csv
+orangecontrib/subgroups/datasets/students.tab
+orangecontrib/subgroups/mrsd/DEV_NOTES.md
+orangecontrib/subgroups/mrsd/README.md
+orangecontrib/subgroups/mrsd/__init__.py
+orangecontrib/subgroups/mrsd/classifiers.py
+orangecontrib/subgroups/mrsd/clustering.py
+orangecontrib/subgroups/mrsd/common.py
+orangecontrib/subgroups/mrsd/core.py
+orangecontrib/subgroups/mrsd/data_loader.py
+orangecontrib/subgroups/mrsd/divergence.py
+orangecontrib/subgroups/mrsd/inference.py
+orangecontrib/subgroups/mrsd/mrsd_ensemble.py
+orangecontrib/subgroups/mrsd/mrsd_mi.py
+orangecontrib/subgroups/mrsd/mrsd_mv.py
+orangecontrib/subgroups/mrsd/mrsd_rules.py
+orangecontrib/subgroups/mrsd/mrsdc2_ranking.py
+orangecontrib/subgroups/mrsd/mrsdc2_semisupervised.py
+orangecontrib/subgroups/mrsd/mrsdc2_swapping.py
+orangecontrib/subgroups/mrsd/mrsdcca.py
+orangecontrib/subgroups/mrsd/mrsdemm.py
+orangecontrib/subgroups/mrsd/mrsdhc.py
+orangecontrib/subgroups/mrsd/mrsdpt.py
+orangecontrib/subgroups/mrsd/rules.py
+orangecontrib/subgroups/mrsd/subgroups.py
+orangecontrib/subgroups/mrsd/synthetic.py
+orangecontrib/subgroups/widgets/__init__.py
+orangecontrib/subgroups/widgets/inspectors.py
+orangecontrib/subgroups/widgets/owbase.py
+orangecontrib/subgroups/widgets/owloadsubgroups.py
+orangecontrib/subgroups/widgets/owmrsd_c2.py
+orangecontrib/subgroups/widgets/owmrsd_cca.py
+orangecontrib/subgroups/widgets/owmrsd_emm.py
+orangecontrib/subgroups/widgets/owmrsd_ensemble.py
+orangecontrib/subgroups/widgets/owmrsd_hc.py
+orangecontrib/subgroups/widgets/owmrsd_mi.py
+orangecontrib/subgroups/widgets/owmrsd_mv.py
+orangecontrib/subgroups/widgets/owmrsd_pt.py
+orangecontrib/subgroups/widgets/owmrsd_ranking.py
+orangecontrib/subgroups/widgets/owmrsd_rules.py
+orangecontrib/subgroups/widgets/owmrsd_semisupervised.py
+orangecontrib/subgroups/widgets/owmrsd_swapping.py
+orangecontrib/subgroups/widgets/owsavesubgroups.py
+orangecontrib/subgroups/widgets/owsubgroupbrowser.py
+orangecontrib/subgroups/widgets/owsubgroupclassifier.py
+orangecontrib/subgroups/widgets/owsubgroupconsolidation.py
+orangecontrib/subgroups/widgets/owsubgroupdatasets.py
+orangecontrib/subgroups/widgets/owsubgroupoverlap.py
+orangecontrib/subgroups/widgets/owsubgroupprofile.py
+orangecontrib/subgroups/widgets/owvariableroles.py
+orangecontrib/subgroups/widgets/utils.py
+orangecontrib/subgroups/widgets/icons/browser.svg
+orangecontrib/subgroups/widgets/icons/c2.svg
+orangecontrib/subgroups/widgets/icons/category.svg
+orangecontrib/subgroups/widgets/icons/cca.svg
+orangecontrib/subgroups/widgets/icons/classifier.svg
+orangecontrib/subgroups/widgets/icons/consolidation.svg
+orangecontrib/subgroups/widgets/icons/datasets.svg
+orangecontrib/subgroups/widgets/icons/emm.svg
+orangecontrib/subgroups/widgets/icons/ensemble.svg
+orangecontrib/subgroups/widgets/icons/hc.svg
+orangecontrib/subgroups/widgets/icons/load.svg
+orangecontrib/subgroups/widgets/icons/mi.svg
+orangecontrib/subgroups/widgets/icons/mv.svg
+orangecontrib/subgroups/widgets/icons/overlap.svg
+orangecontrib/subgroups/widgets/icons/profile.svg
+orangecontrib/subgroups/widgets/icons/pt.svg
+orangecontrib/subgroups/widgets/icons/ranking.svg
+orangecontrib/subgroups/widgets/icons/roles.svg
+orangecontrib/subgroups/widgets/icons/rules.svg
+orangecontrib/subgroups/widgets/icons/save.svg
+orangecontrib/subgroups/widgets/icons/semisupervised.svg
+orangecontrib/subgroups/widgets/icons/swapping.svg
+tests/__init__.py
+tests/conftest.py
+tests/test_backend.py
+tests/test_params.py
+tests/test_widgets.py
+tools/make_datasets.py
+tools/make_docs.py
+tools/make_screenshots.py
+tools/sync_mrsd.py

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/entry_points.txt

@@ -0,0 +1,8 @@
+[orange.data.io.search_paths]
+subgroups = orangecontrib.subgroups.datasets
+
+[orange.widgets]
+Subgroups = orangecontrib.subgroups.widgets
+
+[orange3.addon]
+subgroups = orangecontrib.subgroups

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/not-zip-safe

@@ -0,0 +1 @@
+

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/requires.txt

@@ -0,0 +1,12 @@
+Orange3>=3.34
+numpy>=1.23
+pandas>=1.5
+scipy>=1.9
+scikit-learn>=1.2
+pyqtgraph
+
+[excel]
+openpyxl
+
+[test]
+pytest

orange3_mrsd-0.1.0/Orange3_MRSD.egg-info/top_level.txt

@@ -0,0 +1,2 @@
+orangecontrib
+tests

orange3_mrsd-0.1.0/PKG-INFO

@@ -0,0 +1,103 @@
+Metadata-Version: 2.4
+Name: Orange3-MRSD
+Version: 0.1.0
+Summary: Orange add-on for discovering subgroups in data with several dependent variables (MRSD).
+Home-page: https://github.com/lan-umek/orange3-mrsd
+Author: Lan Umek
+License: MIT
+Keywords: orange3 add-on,subgroup discovery,multiple responses,MRSD,exceptional model mining
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Information Analysis
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: Orange3>=3.34
+Requires-Dist: numpy>=1.23
+Requires-Dist: pandas>=1.5
+Requires-Dist: scipy>=1.9
+Requires-Dist: scikit-learn>=1.2
+Requires-Dist: pyqtgraph
+Provides-Extra: excel
+Requires-Dist: openpyxl; extra == "excel"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Dynamic: author
+Dynamic: classifier
+Dynamic: description
+Dynamic: description-content-type
+Dynamic: home-page
+Dynamic: keywords
+Dynamic: license
+Dynamic: license-file
+Dynamic: provides-extra
+Dynamic: requires-dist
+Dynamic: requires-python
+Dynamic: summary
+
+# Orange3-MRSD
+
+An [Orange](https://orangedatamining.com) add-on for **Multiple Responses
+Subgroup Discovery (MRSD)**: finding groups of instances, described by the
+independent variables **X**, in which several dependent variables **Y** – or
+their relation to X – behave differently from the rest of the data.
+
+The add-on wraps the MRSD Python package (vendored in
+`orangecontrib/subgroups/mrsd`) and adds widgets for setting variable roles,
+running the algorithms, inspecting, combining, visualising and saving the
+subgroups.
+
+## Widgets (category *Subgroups*)
+
+| | |
+|---|---|
+| **Data** | Subgroup Datasets, Variable Roles |
+| **Algorithms** | MRSD-PT (partitioning and tables), MRSD-HC (hierarchical clustering and classification), MRSD-C2 (classification in both spaces: swapping, semi-supervised, ranking), MRSD-CCA, MRSD-EMM, MRSD-MV (multi-view), MRSD-MI, MRSD-Rules, MRSD-Ensemble |
+| **Results** | Subgroup Browser, Subgroup Consolidation, Subgroup Classifier, Subgroup Profile, Subgroup Overlap, Save Subgroups, Load Subgroups |
+
+Every algorithm widget shows the subgroups with their rule in X and prototype
+in Y, has algorithm-specific views of the selected subgroup (contingency
+table, dendrograms, classifiers, correlations …), accepts Orange learners and
+sends the selected subgroup's rows, membership data and membership model on
+to other Orange widgets.
+
+Documentation: [doc/index.md](doc/index.md) – start with
+[Getting started](doc/getting-started.md).
+
+## Installation
+
+Orange 3.34 or newer is required.
+
+```
+pip install Orange3-MRSD        # into the environment that runs Orange
+```
+
+or from a clone: `pip install -e .`.  Excel export additionally needs
+`openpyxl`.
+
+## Development
+
+```
+python -m pytest tests -q        # backend tests run without Orange; widget tests need Orange and QT_QPA_PLATFORM=offscreen
+python tools/make_docs.py        # regenerate doc/ (parameter tables come from params.py)
+python tools/make_datasets.py    # regenerate the bundled datasets
+python tools/sync_mrsd.py        # copy the MRSD package from ../subgroups into orangecontrib/subgroups/mrsd
+```
+
+* The widgets build their parameter GUI from `orangecontrib/subgroups/params.py`;
+  `tests/test_params.py` fails if a field of an algorithm's `Config` dataclass is
+  not covered or a parameter name clashes with a widget attribute.
+* Orange caches the widget descriptions of an add-on and refreshes them only
+  when `orangecontrib/subgroups/widgets/__init__.py` changes: after changing a
+  widget's inputs, outputs or metadata bump `WIDGET_REGISTRY_STAMP` there.
+
+## Status
+
+The algorithms and the non-GUI part of the add-on are covered by tests.  Most widgets have been tried interactively on Orange 3.40 (Windows); some of
+the newer views (e.g. the dendrograms of MRSD-MV, cell selection in Subgroup
+Overlap) and other Orange versions and platforms have not been checked yet.
+
+## License
+
+MIT – see [LICENSE](LICENSE).

orange3_mrsd-0.1.0/README.md

@@ -0,0 +1,65 @@
+# Orange3-MRSD
+
+An [Orange](https://orangedatamining.com) add-on for **Multiple Responses
+Subgroup Discovery (MRSD)**: finding groups of instances, described by the
+independent variables **X**, in which several dependent variables **Y** – or
+their relation to X – behave differently from the rest of the data.
+
+The add-on wraps the MRSD Python package (vendored in
+`orangecontrib/subgroups/mrsd`) and adds widgets for setting variable roles,
+running the algorithms, inspecting, combining, visualising and saving the
+subgroups.
+
+## Widgets (category *Subgroups*)
+
+| | |
+|---|---|
+| **Data** | Subgroup Datasets, Variable Roles |
+| **Algorithms** | MRSD-PT (partitioning and tables), MRSD-HC (hierarchical clustering and classification), MRSD-C2 (classification in both spaces: swapping, semi-supervised, ranking), MRSD-CCA, MRSD-EMM, MRSD-MV (multi-view), MRSD-MI, MRSD-Rules, MRSD-Ensemble |
+| **Results** | Subgroup Browser, Subgroup Consolidation, Subgroup Classifier, Subgroup Profile, Subgroup Overlap, Save Subgroups, Load Subgroups |
+
+Every algorithm widget shows the subgroups with their rule in X and prototype
+in Y, has algorithm-specific views of the selected subgroup (contingency
+table, dendrograms, classifiers, correlations …), accepts Orange learners and
+sends the selected subgroup's rows, membership data and membership model on
+to other Orange widgets.
+
+Documentation: [doc/index.md](doc/index.md) – start with
+[Getting started](doc/getting-started.md).
+
+## Installation
+
+Orange 3.34 or newer is required.
+
+```
+pip install Orange3-MRSD        # into the environment that runs Orange
+```
+
+or from a clone: `pip install -e .`.  Excel export additionally needs
+`openpyxl`.
+
+## Development
+
+```
+python -m pytest tests -q        # backend tests run without Orange; widget tests need Orange and QT_QPA_PLATFORM=offscreen
+python tools/make_docs.py        # regenerate doc/ (parameter tables come from params.py)
+python tools/make_datasets.py    # regenerate the bundled datasets
+python tools/sync_mrsd.py        # copy the MRSD package from ../subgroups into orangecontrib/subgroups/mrsd
+```
+
+* The widgets build their parameter GUI from `orangecontrib/subgroups/params.py`;
+  `tests/test_params.py` fails if a field of an algorithm's `Config` dataclass is
+  not covered or a parameter name clashes with a widget attribute.
+* Orange caches the widget descriptions of an add-on and refreshes them only
+  when `orangecontrib/subgroups/widgets/__init__.py` changes: after changing a
+  widget's inputs, outputs or metadata bump `WIDGET_REGISTRY_STAMP` there.
+
+## Status
+
+The algorithms and the non-GUI part of the add-on are covered by tests.  Most widgets have been tried interactively on Orange 3.40 (Windows); some of
+the newer views (e.g. the dendrograms of MRSD-MV, cell selection in Subgroup
+Overlap) and other Orange versions and platforms have not been checked yet.
+
+## License
+
+MIT – see [LICENSE](LICENSE).

orange3_mrsd-0.1.0/doc/concepts.md

@@ -0,0 +1,74 @@
+# Concepts, measures and inference
+
+## The task
+
+The data has independent (descriptive) variables **X** and several dependent
+variables (responses) **Y**, numeric or categorical.  A **subgroup** is a set
+of instances that is (a) describable in X and (b) interesting in Y: the
+dependent variables, or their relation to X, behave differently inside the
+subgroup than in the rest of the data.  This is **Multiple Responses Subgroup
+Discovery (MRSD)**.
+
+## What the algorithms look for
+
+| algorithm | name | a subgroup is … |
+|---|---|---|
+| MRSD-PT | partitioning and tables | a cell of a contingency table of a partition of X against a partition of Y that holds more instances than expected |
+| MRSD-HC | hierarchical clustering and classification | a cluster of Y that a classifier recognises from X |
+| MRSD-C2 | classification in both spaces | a set whose membership is predictable from X and from Y |
+| MRSD-Rules | rules (CN2-SD) | a rule in X with a shifted distribution of Y |
+| MRSD-MV | multi-view | a group coherent in both views X and Y |
+| MRSD-EMM | exceptional model mining | a rule in X under which the X–Y model differs |
+| MRSD-CCA | canonical correlation | a cluster with a different canonical correlation |
+| MRSD-MI | mutual information | a rule in X under which X and Y are more dependent |
+| MRSD-Ensemble | consensus | a group supported by several of the above |
+
+The first five look for a different *distribution* of Y; EMM, CCA and MI look
+for a different *relation* between X and Y.  A subgroup of the second kind
+need not differ in the means of Y at all.
+
+## Description of a subgroup
+
+* **Rule / description (X)** – for rule-based algorithms the rule that defines
+  the subgroup; for the others a decision-tree approximation, with its
+  precision and recall with respect to the subgroup in the *Measures* tab.
+* **Prototype (Y)** – the means of the numeric and the modal levels (with
+  their shares) of the categorical dependent variables in the subgroup.
+* **Classifier** – any Orange learner can describe the membership instead of
+  a rule (*Learner* input, *Model* outputs).
+
+## Quality
+
+Each algorithm has its own quality score, comparable only within that
+algorithm: an AUC for HC and C2, test-statistic-like scores for PT, Rules,
+EMM, CCA, MI, bounded mixtures for MV.  To compare subgroups of different
+algorithms use the **common measures** of *Subgroup Consolidation* (Y
+distinctness, X describability, relation exceptionality), which are computed
+from the rows of a subgroup in the same way for all algorithms.
+
+## Inference
+
+The null hypothesis is that X and Y are independent.  Rows of the view that
+was not used to define the candidates are permuted.  When candidates are
+defined from both views, a test of a fixed row set is anti-conservative, so
+the whole search is re-run on permuted data.  Reported are the raw p-value,
+the family-wise adjusted p-value (covering every candidate the search
+examined) and the Benjamini–Hochberg value.
+
+Permutation inference is optional and off by default except in MRSD-PT.
+
+## What the numbers do not tell
+
+All measures, including cross-validated AUCs and the tests in the *Prototype*
+tab, are computed on the data in which the subgroup was found.  They compare
+subgroups and guard against some chance findings, but they do not estimate
+how well a subgroup would replicate in new data.  For that, set data aside
+before the analysis and examine the found rules or membership models on it.
+
+## Redundancy
+
+Algorithms return many overlapping subgroups.  Within an algorithm they are
+pruned by Jaccard similarity (*Redundancy threshold*).  Across algorithms
+*Subgroup Consolidation* applies the cleaning procedure: take the best
+subgroup by a quality q, remove all subgroups at least s_min similar to it,
+repeat.

orange3_mrsd-0.1.0/doc/getting-started.md

@@ -0,0 +1,79 @@
+# Getting started
+
+## Installation
+
+The add-on needs Orange 3.34 or newer.  Install it into the Python
+environment that runs Orange:
+
+```
+pip install Orange3-MRSD
+```
+
+or, from a clone of the repository, `pip install -e .`.  With the standalone
+Orange installer on Windows use Orange's own Python, e.g.
+
+```
+"%LOCALAPPDATA%\Programs\Orange\python.exe" -m pip install Orange3-MRSD
+```
+
+Restart Orange; the widgets are in the category **Subgroups**.  Excel export
+needs `openpyxl`.
+
+## A first workflow
+
+1. **Subgroup Datasets** → choose *Planted rule*.  The data has three
+   independent variables (age, edu, inc) and two dependent variables (y1, y2);
+   in the subgroup `edu = high AND age ≥ 50` both dependent variables are
+   shifted.
+2. Connect it to **MRSD-Rules** and press *Run*.  The first rule in the table
+   should be close to the planted one.  Select a row and look at the tabs
+   *Rule*, *Prototype* and *Measures*.
+3. Connect the *Subgroups* output to **Subgroup Browser** to filter, rename
+   and select subgroups, and its *Annotated Data* output to Orange's *Scatter
+   Plot*; colour the points by a subgroup column and compare with the meta
+   column `planted`.
+
+## Your own data
+
+The algorithm widgets take the **independent variables X from the attributes**
+and the **dependent variables Y from the class variables** of the data;
+meta attributes are ignored.  Set the roles with **Variable Roles** (or
+Orange's *Select Columns*).  In a `.tab` file several columns can be flagged
+`class`, and then *File* sets the roles by itself.
+
+## Combining algorithms
+
+Connect several algorithm widgets to **Subgroup Consolidation**.  It computes
+measures that are comparable across algorithms and removes near-duplicates;
+its output goes to the Browser, Profile, Overlap and Save widgets.
+
+```
+Data ─► MRSD-Rules ─┐
+     ├► MRSD-HC   ──┼─► Subgroup Consolidation ─► Subgroup Browser ─► Subgroup Profile / Overlap / Save
+     └► MRSD-PT   ──┘
+```
+
+## Describing a subgroup with a model
+
+Connect an Orange learner (Tree, Naive Bayes, Logistic Regression) to the
+*Learner* input of an algorithm widget.  For the selected subgroup the widget
+then sends the fitted membership model: connect *Model (X)* to *Tree Viewer*
+or to *Nomogram* (Nomogram shows Naive Bayes and logistic regression models;
+add *Membership Data (X)* → *Data* to see individual instances).
+
+## Speed
+
+* Permutation tests are off by default except in MRSD-PT.  Search-level
+  permutations repeat the whole search; start with a small number.
+* Classification-based methods (HC, C2, MV, CCA) are fast with a decision
+  tree and slow with a random forest or SVM.
+* A widget placed before an update keeps its saved settings; new defaults
+  apply to newly placed widgets.
+
+## Troubleshooting
+
+Start Orange from a terminal (`python -m Orange.canvas`); the add-on writes
+its progress and any error there in lines beginning with `[MRSD`.  If a widget
+lacks inputs or outputs described here, Orange is using a cached description
+of the add-on: close Orange, delete the files `*.pck` in Orange's cache folder
+(`%LOCALAPPDATA%\Orange\cache\<version>` on Windows) and start it again.