MCPower 0.4.0__tar.gz → 0.4.2__tar.gz

{mcpower-0.4.0 → mcpower-0.4.2}/.github/workflows/release.yml

@@ -69,7 +69,7 @@ jobs:
         CIBW_MANYLINUX_X86_64_IMAGE: manylinux_2_28
         CIBW_MANYLINUX_AARCH64_IMAGE: manylinux_2_28
         CIBW_TEST_REQUIRES: "pytest numpy pandas matplotlib scipy statsmodels"
-        CIBW_TEST_COMMAND: "pytest {project}/tests -v -x --ignore={project}/tests/backend/ -m \"not slow\""
+        CIBW_TEST_COMMAND: "pytest {project}/tests -v -x --ignore={project}/tests/backend/ -m \"not slow and not lme\""
 
     - name: Verify native extension in wheels
       shell: bash

{mcpower-0.4.0 → mcpower-0.4.2}/.github/workflows/test-release.yml

@@ -34,7 +34,7 @@ jobs:
         CIBW_MANYLINUX_X86_64_IMAGE: manylinux_2_28
         CIBW_MANYLINUX_AARCH64_IMAGE: manylinux_2_28
         CIBW_TEST_REQUIRES: "pytest numpy pandas matplotlib scipy statsmodels"
-        CIBW_TEST_COMMAND: "pytest {project}/tests -v -x --ignore={project}/tests/backend/ -m \"not slow\""
+        CIBW_TEST_COMMAND: "pytest {project}/tests -v -x --ignore={project}/tests/backend/ -m \"not slow and not lme\""
 
     - name: Verify native extension in wheels
       shell: bash

{mcpower-0.4.0 → mcpower-0.4.2}/.github/workflows/tests.yml

@@ -37,12 +37,12 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install pytest pytest-cov
-        pip install pandas joblib matplotlib statsmodels
+        pip install pandas joblib matplotlib
         pip install numba || echo "Numba not available for Python ${{ matrix.python-version }}, will use Python fallback"
         pip install -e .
 
     - name: Run tests with coverage
-      run: pytest tests/ -v --tb=short --cov=mcpower --cov-report=xml -m "not slow"
+      run: pytest tests/ -v --tb=short --cov=mcpower --cov-report=xml -m "not slow and not lme"
 
     - name: Upload coverage to Codecov
       uses: codecov/codecov-action@v4
@@ -98,7 +98,7 @@ jobs:
     - name: Test Python fallback mode
       run: |
         python -c "import mcpower; print('Python fallback mode working')"
-        pytest tests/ -v --tb=short -m "not slow"
+        pytest tests/ -v --tb=short -m "not slow and not lme"
 
   test-performance:
     runs-on: ubuntu-latest

mcpower-0.4.2/CHANGELOG.md

@@ -0,0 +1,130 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+## [0.4.2] - 2026-02-17
+
+### Major changes
+- **Custom LME solver**: Replaced statsmodels MixedLM with a custom profiled-deviance REML solver (Python + C++ + JIT backends), eliminating the statsmodels runtime dependency for mixed models
+- **C++ LME backend**: LME analysis exposed via PyBind11/Eigen (`mcpower_native.lme_analysis`) for maximum throughput
+- **Post-hoc pairwise comparisons**: `target_test="group[1] vs group[2]"` syntax for contrasts between factor levels
+- **Tukey HSD correction**: New `correction="tukey"` option for pairwise comparisons using studentized range critical values
+
+### Minor changes
+- `target_test` default changed from `"overall"` to `"all"` (overall F-test + all individual fixed effects)
+- `target_test` now supports keyword expansion (`"all"`, `"all-posthoc"`), exclusion prefixes (`"-overall"`), and comma-separated combinations
+- Renamed optional dependency extra from `[mixed]` to `[lme]`
+- Formatter handles `NaN` power values gracefully (displays `"-"` instead of crashing)
+- README rewritten with clearer examples and MCPower GUI cross-link
+
+### Technical
+- New `mcpower/stats/` package: statistical routines (`ols.py`, `lme_solver.py`, `mixed_models.py`, `data_generation.py`) extracted from `utils/`
+- Reclassified several `@pytest.mark.slow` tests to run in the default (non-slow) suite
+- Removed old test files and `scripts/test_quick.py`; moved `LME_BENCHMARK_TESTS.md` to `docs/`
+
+### Experimental
+- **Random slopes**: Support for `(1 + x|group)` syntax with configurable slope variance and slope-intercept correlation
+- **Nested random effects**: Support for `(1|A/B)` syntax with automatic expansion to parent and child grouping terms
+
+## [0.4.1] - 2026-02-14
+
+### Minor changes
+- README badges now link to their respective pages
+- Added MCPower GUI desktop application section to README
+
+### Technical
+- Reclassified several backend and mixed-model tests from `@pytest.mark.slow` to default suite
+- Moved slow markers to individual test methods for finer-grained control
+
+## [0.4.0] - 2026-02-12
+
+Releasing pipeline fixed
+
+## [0.4.0b0] - 2026-02-12
+
+The package has been almost entirely rebuilt from the ground up.
+
+### Major changes
+- **C++ native backend**: New PyBind11/Eigen-based compute backend for OLS and data generation (~3x speedup), removed Numba AOT
+- **Multi-backend architecture**: Automatic selection between C++ native, Numba JIT, and pure Python backends with graceful fallback
+- **Mixed-effects models**: Random intercept models via `(1|group)` syntax with statsmodels MixedLM
+- **Build system migration**: Moved from setuptools to scikit-build-core with CMake for C++ compilation
+- **Python 3.14 support**
+- **Lookup tables reduced and reorganized**: removed lookup tables from OLS, precomputed critical values added
+
+### Minor changes
+- `joblib` is now a core dependency; `statsmodels` moved to optional `[lme]` extra
+- Progress callback system for GUI/notebook integration
+
+### Technical
+- Complete code reorganization: monolithic `base.py` split into modular structure (`mcpower/core/`, `mcpower/backends/`, `mcpower/model.py`)
+- New CI/CD pipeline: separate workflows for stable releases (PyPI) and pre-releases (TestPyPI)
+- Comprehensive test suite overhaul (650 tests across unit, integration, spec, and mixed-model categories)
+- Pre-commit hooks with ruff for linting and formatting
+- Removed legacy `setup.py`, `pytest.ini`, and `paper/` directory
+
+## [0.3.3] - 2025-07-27
+
+### Technical
+- Wheels for Python 3.13
+
+## [0.3.2] - 2025-07-27
+
+### Major changes
+- Factors are now supported
+- The workflow requires that `set_variable_types` be set before correlation and effects
+
+## [0.3.1] - 2025-07-17
+
+### Major changes
+- Added weekly update checker
+
+### Minor changes
+- Small changes in documentation
+
+## [0.3.0] - 2025-07-16
+
+### Major changes
+- AOT compiled files added
+- PyPI release
+
+### Minor changes
+- Small changes in documentation, added fallback to compilation
+
+## [0.2.4] - 2025-07-16
+- Test PyPI release
+
+## [0.2.3] - 2025-07-16
+- Test PyPI release
+
+## [0.2.2] - 2025-07-11
+
+### Major changes
+- Moved seed setting from method arguments to `model.set_seed()` for better control
+- Added public method to customize scenario configurations
+- Added support for 3-way and n-way interactions in formulas
+
+### Minor changes
+- Improved lookup table performance with extended range and faster fallbacks
+- Fixed import error when uploading user data
+- Fixed minor cache creation error
+- Adjusted vulnerability thresholds from 30% to 20% power drop
+- Added warning for inflated Type I error risk
+
+### Technical
+- Added comprehensive test suite
+- Improved code documentation with clearer comments and global variables
+- Clarified correlation behavior during variable transformations
+
+## [0.2.1] - 2025-06-26
+
+### Minor changes
+- Fixed import statements to use relative imports instead of absolute imports
+- Added changelog
+- Extended docstrings
+
+## [0.2.0] - 2025-06-26
+- Initial release
+
+## [0.1.0] - 2025-06-07
+- Proof of concept release

{mcpower-0.4.0 → mcpower-0.4.2}/CMakeLists.txt

@@ -6,11 +6,32 @@ set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
-# Find pybind11 — optional so sdist installs without a compiler still succeed
-find_package(pybind11 CONFIG)
+# Find pybind11 — try CONFIG first, then query the Python module as fallback
+find_package(pybind11 CONFIG QUIET)
 if(NOT pybind11_FOUND)
-    message(WARNING "pybind11 not found — building Python-only package (no C++ backend)")
-    return()
+    # Fallback: ask the pybind11 Python module where its cmake files live
+    # (scikit-build-core may not always add site-packages to CMAKE_PREFIX_PATH on Windows)
+    find_package(Python COMPONENTS Interpreter QUIET)
+    if(Python_FOUND)
+        execute_process(
+            COMMAND "${Python_EXECUTABLE}" -m pybind11 --cmakedir
+            OUTPUT_VARIABLE _pybind11_cmake_dir
+            OUTPUT_STRIP_TRAILING_WHITESPACE
+            ERROR_QUIET
+            RESULT_VARIABLE _pybind11_result
+        )
+        if(_pybind11_result EQUAL 0 AND _pybind11_cmake_dir)
+            find_package(pybind11 CONFIG REQUIRED PATHS "${_pybind11_cmake_dir}" NO_DEFAULT_PATH)
+        endif()
+    endif()
+endif()
+if(NOT pybind11_FOUND)
+    if(DEFINED ENV{CIBUILDWHEEL})
+        message(FATAL_ERROR "pybind11 not found in CI — cannot build wheel without native extension")
+    else()
+        message(WARNING "pybind11 not found — building Python-only package (no C++ backend)")
+        return()
+    endif()
 endif()
 
 # Try to find Eigen3, make it optional
@@ -47,6 +68,7 @@ endif()
 set(MCPOWER_SOURCES
     cpp/src/ols.cpp
     cpp/src/data_generation.cpp
+    cpp/src/lme_solver.cpp
     cpp/src/bindings.cpp
 )
 
@@ -78,6 +100,7 @@ elseif(MSVC)
         /fp:fast
         /DNDEBUG
     )
+    target_compile_definitions(mcpower_native PRIVATE _USE_MATH_DEFINES)
 endif()
 
 # OpenMP support (optional)

{mcpower-0.4.0 → mcpower-0.4.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: MCPower
-Version: 0.4.0
+Version: 0.4.2
 Summary: Monte Carlo Power Analysis for Statistical Models
 Keywords: power analysis,statistics,monte carlo,linear regression
 Author-Email: =?utf-8?q?Pawe=C5=82_Lenartowicz?= <pawellenartowicz@europe.com>
@@ -28,8 +28,8 @@ Requires-Dist: scipy>=1.11.0
 Requires-Dist: joblib>=1.3.0
 Provides-Extra: jit
 Requires-Dist: numba>=0.61.0; extra == "jit"
-Provides-Extra: mixed
-Requires-Dist: statsmodels>=0.14.0; extra == "mixed"
+Provides-Extra: lme
+Requires-Dist: statsmodels>=0.14.0; extra == "lme"
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
@@ -45,11 +45,11 @@ Requires-Dist: pytest>=7.0.0; extra == "all"
 Requires-Dist: pytest-cov>=4.0.0; extra == "all"
 Description-Content-Type: text/markdown
 
-![Tests](https://github.com/pawlenartowicz/MCPower/workflows/Tests/badge.svg)
-![PyPI](https://img.shields.io/pypi/v/MCPower)
-![Python](https://img.shields.io/pypi/pyversions/mcpower)
-![License: GPL v3](https://img.shields.io/badge/License-GPLv3-green.svg)
-[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.16502734.svg)](https://doi.org/10.5281/zenodo.16502734)
+[![Tests](https://github.com/pawlenartowicz/MCPower/workflows/Tests/badge.svg)](https://github.com/pawlenartowicz/MCPower/actions)
+[![PyPI](https://img.shields.io/pypi/v/MCPower)](https://pypi.org/project/MCPower/)
+[![Python](https://img.shields.io/pypi/pyversions/mcpower)](https://pypi.org/project/MCPower/)
+[![License: GPL v3](https://img.shields.io/badge/License-GPLv3-green.svg)](https://www.gnu.org/licenses/gpl-3.0)
+[![DOI](https://img.shields.io/badge/DOI-10.5281%2Fzenodo.16502734-blue)](https://doi.org/10.5281/zenodo.16502734)
 
 ```
 ███╗   ███╗  ██████╗ ██████╗ 
@@ -64,16 +64,23 @@ Description-Content-Type: text/markdown
 
 **Simple Monte Carlo power analysis for complex models.** Find the sample size you need or check if your study has enough power - even with complex models that traditional power analysis can't handle.
 
+## Desktop Application
+
+It it Python package, prefer a graphical interface? **[MCPower GUI](https://github.com/pawlenartowicz/mcpower-gui)** is a standalone desktop app — no Python installation required. Download ready-to-run executables for Windows, Linux, and macOS from the [releases page](https://github.com/pawlenartowicz/mcpower-gui/releases/latest).
+
 ## Why MCPower?
 
-**Traditional power analysis breaks down** with interactions, correlated predictors, categorical variables, or non-normal data. MCPower uses simulation instead of formulas - it generates thousands of datasets exactly like yours, then sees how often your analysis finds real effects.
+**Traditional power analysis breaks down** with interactions, correlated predictors, categorical variables, or non-normal data. MCPower uses simulation instead of formulas — it generates thousands of datasets exactly like yours, then sees how often your analysis finds real effects.
+
+**Just write your formula.** Define your model the way you'd write it in R — `outcome = treatment + covariate + treatment*covariate`. MCPower parses the formula, sets up the simulation, and handles interactions, factor coding, and dummy variables automatically. You focus on the research question, not the mechanics. More model types (logistic regression, ANOVA) are on the way.
+
+**Test your assumptions with scenarios.** Real studies rarely match textbook conditions — effect sizes may be smaller than expected, distributions may be skewed, or variance may not be constant. Turn on `scenarios=True` and MCPower automatically tests your power under optimistic, realistic, and worst-case conditions. Instead of a single number, you get a range that shows how sensitive your design is to violated assumptions — so you plan for reality, not just the best case.
+
+**Use your own data.** Upload a CSV and MCPower auto-detects variable types (continuous, binary, or categorical), preserves real distributions, and handles correlations between predictors. No need to overthink whether your data is normal, skewed, or categorical — just upload it and MCPower samples from the empirical distribution. This is especially useful when you have pilot data or a related dataset and want your power analysis to reflect actual conditions rather than idealized ones.
 
-✅ **Works with complexity**: Interactions, correlations, factors, any distribution  
-✅ **R-style formulas**: `outcome = treatment + covariate + treatment*covariate`  
-✅ **Categorical variables**: Multi-level factors automatically handled  
-✅ **Two simple commands**: Find sample size or check power  
-✅ **Scenario analysis**: Test robustness under realistic conditions  
-✅ **Minimal math required**: Just specify your model and effects
+✅ **Works with complexity**: Interactions, correlations, multi-level factors, any distribution
+✅ **Two simple commands**: `find_sample_size()` or `find_power()` — that's the entire API
+✅ **Minimal math required**: Just specify your model formula and expected effect sizes
 
 ## Get Started in 2 Minutes
 
@@ -91,10 +98,10 @@ pip install --upgrade mcpower
 ```python
 
 # 0. Import installed package
-import mcpower
+from mcpower import MCPower
 
 # 1. Define your model (just like R)
-model = mcpower.MCPower("satisfaction = treatment + motivation")
+model = MCPower("satisfaction = treatment + motivation")
 
 # 2. Set effect sizes (how big you expect effects to be)
 model.set_effects("treatment=0.5, motivation=0.3")
@@ -175,10 +182,10 @@ model.set_effects("treatment=0.5, age=0.3, income=0.2")
 
 ### Randomized Controlled Trial
 ```python
-import mcpower
+from mcpower import MCPower
 
 # RCT with treatment + control variables
-model = mcpower.MCPower("outcome = treatment + age + baseline_score")
+model = MCPower("outcome = treatment + age + baseline_score")
 model.set_effects("treatment=0.6, age=0.2, baseline_score=0.8")
 model.set_variable_type("treatment=binary")  # 0/1 treatment
 
@@ -189,10 +196,10 @@ model.find_sample_size(target_test="treatment", from_size=100, to_size=500,
 
 ### A/B Test with Interaction
 ```python
-import mcpower
+from mcpower import MCPower
 
 # Test if treatment effect depends on user type
-model = mcpower.MCPower("conversion = treatment + user_type + treatment*user_type")
+model = MCPower("conversion = treatment + user_type + treatment*user_type")
 model.set_effects("treatment=0.4, user_type=0.3, treatment:user_type=0.5")
 model.set_variable_type("treatment=binary, user_type=binary")
 
@@ -202,10 +209,10 @@ model.find_power(sample_size=400, target_test="treatment:user_type", scenarios=T
 
 ### Multi-Group Study with Categorical Variables
 ```python
-import mcpower
+from mcpower import MCPower
 
 # Study with 3 treatment groups and 4 education levels
-model = mcpower.MCPower("wellbeing = treatment + education + age")
+model = MCPower("wellbeing = treatment + education + age")
 model.set_variable_type("treatment=(factor,3), education=(factor,4)")
 
 # Set effects for each factor level (vs. reference level 1)
@@ -217,10 +224,10 @@ model.find_sample_size(target_test="treatment[2], treatment[3]", scenarios=True)
 
 ### Survey with Correlated Predictors
 ```python
-import mcpower
+from mcpower import MCPower
 
 # Predictors are often correlated in real data
-model = mcpower.MCPower("wellbeing = income + education + social_support")
+model = MCPower("wellbeing = income + education + social_support")
 model.set_effects("income=0.4, education=0.3, social_support=0.6")
 model.set_correlations("corr(income, education)=0.5, corr(income, social_support)=0.3")
 
@@ -246,7 +253,7 @@ model.set_variable_type("condition=(factor,0.2,0.5,0.3)")
 ### Working with Factors (Categorical Variables)
 ```python
 # Factors automatically create dummy variables
-model = mcpower.MCPower("outcome = treatment + education")
+model = MCPower("outcome = treatment + education")
 model.set_variable_type("treatment=(factor,3), education=(factor,4)")
 
 # Set effects for specific levels (level 1 is always reference)
@@ -264,7 +271,7 @@ import pandas as pd
 data = pd.read_csv("my_data.csv")
 
 # Upload with automatic type detection
-model = mcpower.MCPower("mpg = hp + wt + cyl")
+model = MCPower("mpg = hp + wt + cyl")
 model.upload_data(data[["hp", "wt", "cyl"]])
 model.set_effects("hp=0.5, wt=0.3, cyl[2]=0.2, cyl[3]=0.4")
 model.find_power(sample_size=100)
@@ -308,13 +315,28 @@ model.upload_data(
 ```python
 # Testing multiple effects? Control false positives
 model.find_power(
-    sample_size=200, 
+    sample_size=200,
     target_test="treatment,covariate,treatment:covariate",
     correction="Benjamini-Hochberg",
     scenarios=True  # Test robustness too!
 )
 ```
 
+### Post-Hoc Pairwise Comparisons (Tukey HSD)
+```python
+# Compare specific factor levels with Tukey correction
+model = MCPower("outcome = group + covariate")
+model.set_variable_type("group=(factor,3)")
+model.set_effects("group[2]=0.4, group[3]=0.6, covariate=0.3")
+
+# Use "vs" syntax for pairwise comparisons + correction="tukey"
+model.find_power(
+    sample_size=150,
+    target_test="group[0] vs group[1], group[0] vs group[2]",
+    correction="tukey"
+)
+```
+
 ### Test the single violation of assumptions.
 ```python
 # Customize how much "messiness" to add in scenarios
@@ -325,27 +347,37 @@ model.set_heteroskedasticity(0.15)  # Violation of equal variance assumption
 model.find_sample_size(target_test="treatment", scenarios=False)
 ```
 
-### Mixed-Effects Models (Random Intercept)
+### Mixed-Effects Models (Experimental)
 ```python
-import mcpower
+from mcpower import MCPower
 
-# Define a model with a random intercept for clustered data
-model = mcpower.MCPower("satisfaction ~ treatment + motivation + (1|school)")
+# Random intercept — clustered data
+model = MCPower("satisfaction ~ treatment + motivation + (1|school)")
 model.set_cluster("school", ICC=0.2, n_clusters=20)
 model.set_effects("treatment=0.5, motivation=0.3")
 model.set_variable_type("treatment=binary")
-
-# Allow some convergence failures (common with mixed models)
-model.set_max_failed_simulations(0.10)  # Up to 10% failures tolerated
-
-# Total sample_size is split across clusters: 1000 / 20 = 50 per cluster
+model.set_max_failed_simulations(0.10)
+model.find_power(sample_size=1000)  # 1000/20 = 50 per cluster
+
+# Random slopes — effect varies across clusters
+model = MCPower("y ~ x1 + (1 + x1|school)")
+model.set_cluster("school", ICC=0.2, n_clusters=20,
+                   random_slopes=["x1"], slope_variance=0.1,
+                   slope_intercept_corr=0.3)
+model.set_effects("x1=0.5")
+model.set_max_failed_simulations(0.30)
 model.find_power(sample_size=1000)
+
+# Nested random effects — students in classrooms in schools
+model = MCPower("y ~ treatment + (1|school/classroom)")
+model.set_cluster("school", ICC=0.15, n_clusters=10)
+model.set_cluster("classroom", ICC=0.10, n_per_parent=3)  # 3 classrooms per school
+model.set_effects("treatment=0.5")
+model.set_max_failed_simulations(0.30)
+model.find_power(sample_size=1500)
 ```
 
-**Constraints:**
-- Only random intercepts `(1|group)` — random slopes are not yet supported
-- At least 25 observations per cluster and 10 observations per model parameter
-- ICC must be between 0.1 and 0.9 (or exactly 0)
+See the [Mixed-Effects Models wiki page](https://github.com/pawlenartowicz/MCPower/wiki/Mixed-Effects-Models) for detailed documentation on all model types, parameters, and design recommendations.
 
 ### More precision
 ```python
@@ -392,8 +424,11 @@ model.find_power(sample_size=200, progress_callback=False)
 | **Factor variables** | **`model.set_variable_type("var=(factor,3)")`** |
 | **Factor effects** | **`model.set_effects("var[2]=0.5, var[3]=0.7")`** |
 | Correlated predictors | `model.set_correlations("corr(var1, var2)=0.4")` |
-| Multiple testing correction | Add `correction="FDR"`, `"Holm"`, or `"Bonferroni"`|
-| Mixed model (random intercept) | `MCPower("y ~ x + (1\|group)")` + `model.set_cluster(...)` |
+| Multiple testing correction | Add `correction="FDR"`, `"Holm"`, `"Bonferroni"`, or `"Tukey"`|
+| Post-hoc pairwise comparison | `target_test="group[0] vs group[1]"` with `correction="tukey"` |
+| Mixed model (random intercept) | `MCPower("y ~ x + (1\|group)")` + `model.set_cluster(...)` (experimental) |
+| Random slopes | `MCPower("y ~ x + (1+x\|group)")` + `set_cluster(..., random_slopes=["x"], slope_variance=0.1)` (experimental) |
+| Nested random effects | `MCPower("y ~ x + (1\|A/B)")` + two `set_cluster()` calls (experimental) |
 | Reproducible results | `model.set_seed(42)` |
 | Get results as dict | Add `return_results=True` to either method |
 | Stricter significance | `model.set_alpha(0.01)` |
@@ -465,7 +500,7 @@ model.set_variable_type("""
 ### Factor Variables in Detail
 ```python
 # Factor variables are categorical with multiple levels
-model = mcpower.MCPower("outcome = treatment + education")
+model = MCPower("outcome = treatment + education")
 
 # Create factors
 model.set_variable_type("treatment=(factor,3), education=(factor,4)")
@@ -539,9 +574,23 @@ model.set_correlations("(x1, x2)=0.3, (x1, x3)=-0.2")
 - Python ≥ 3.10
 - NumPy, SciPy, matplotlib, Pandas, joblib
 - C++ compiler (automatically used during install for native backend; falls back to Python if unavailable)
-- statsmodels (optional, for mixed-effects models — install with `pip install mcpower[mixed]`)
+- statsmodels (optional, for mixed-effects models — install with `pip install mcpower[lme]`)
 - Numba (optional, for JIT compilation fallback — install with `pip install mcpower[JIT]`)
 
+
+## Documentation
+
+Full documentation is available on the **[MCPower Wiki](https://github.com/pawlenartowicz/MCPower/wiki)**, including:
+
+- [Quick Start](https://github.com/pawlenartowicz/MCPower/wiki/Quick-Start)
+- [Model Specification](https://github.com/pawlenartowicz/MCPower/wiki/Model-Specification)
+- [Variable Types](https://github.com/pawlenartowicz/MCPower/wiki/Variable-Types)
+- [Effect Sizes](https://github.com/pawlenartowicz/MCPower/wiki/Effect-Sizes)
+- [Mixed-Effects Models](https://github.com/pawlenartowicz/MCPower/wiki/Mixed-Effects-Models) (random intercepts, slopes, nested effects)
+- [ANOVA & Post-Hoc Tests](https://github.com/pawlenartowicz/MCPower/wiki/ANOVA-and-Post-Hoc-Tests)
+- [Scenario Analysis](https://github.com/pawlenartowicz/MCPower/wiki/Scenario-Analysis)
+- [API Reference](https://github.com/pawlenartowicz/MCPower/wiki/API-Reference)
+
 ## Need Help?
 
 - **Issues**: [GitHub Issues](https://github.com/pawlenartowicz/MCPower/issues)
@@ -552,7 +601,7 @@ model.set_correlations("(x1, x2)=0.3, (x1, x3)=-0.2")
 - ✅ Scenarios, robustness analysis
 - ✅ Factor variables (categorical predictors)
 - ✅ C++ native backend (pybind11 + Eigen, 3x speedup)
-- 🚧 Mixed Effects Models (partly implemented — random intercept only)
+- ⚠️ Mixed Effects Models (random intercepts, random slopes, nested effects) — experimental
 - 🚧 Logistic Regression (coming soon)
 - 🚧 ANOVA (coming soon)
 - 🚧 Guide about methods, corrections (coming soon)