KNIME® — UTD Statistical Analysis Extension

Developed by Ahmed Elghazi and Rabih Neouchi at the University of Texas at Dallas.

This KNIME Python extension provides a suite of statistical analysis nodes for hypothesis testing, variance analysis, and regression diagnostics — packaged for seamless use inside KNIME Analytics Platform.

---

Contents

.
├── icons/
│   ├── utd.png                          # Extension category icon
│   ├── curve.jpg                        # Normality Tests icon
│   ├── post_hoc.jpg                     # Post-Hoc Analysis icon
│   ├── heteroskedasticity.png           # Heteroskedasticity Tests icon
│   ├── factorial.jpg                    # Factorial ANOVA icon
│   ├── manova3.png                      # One-Way MANOVA icon
│   └── rm_anova.png                     # Repeated Measures ANOVA icon
├── src/
│   ├── __init__.py
│   ├── utils.py                         # Shared helpers and parameter definitions
│   ├── normality_node.py
│   ├── normality_tests/
│   ├── post_hoc_node.py
│   ├── post_hoc/
│   ├── heteroskedasticity_node.py
│   ├── heteroskedasticity/
│   ├── factorial_anova_node.py
│   ├── factorial_anova/
│   ├── manova_node.py
│   ├── manova/
│   ├── repeated_measures_anova_node.py
│   └── repeated_measures_anova/
├── golden_tables/                       # Golden reference tables and CSV fixtures
│   ├── data/                            # CSV test datasets
│   ├── generate_normality_golden.py
│   ├── generate_post_hoc_golden.py
│   ├── generate_heteroskedasticity_golden.py
│   ├── generate_factorial_anova_golden.py
│   ├── generate_manova_golden.py
│   └── generate_rm_anova_golden.py
├── workflows/                           # Demo KNIME workflows (.knwf)
├── knime.yml                            # Extension metadata
├── pixi.toml                            # Dependency management
├── pixi.lock                            # Locked dependency versions
├── ruff.toml                            # Code formatting config
├── pytest.ini                           # Pytest configuration
└── LICENSE.TXT

---

Nodes

Statistical Normality Tests

Tests whether one or more numeric columns follow a normal distribution.

• Methods: Anderson-Darling, Cramer-von Mises
• Input: Table with numeric column(s)
• Output: Per-column results — Test Statistic, P-Value, Statistical Decision

Post-Hoc Analysis

Runs a one-way ANOVA and, if significant, identifies which group pairs differ.

• Methods: Tukey HSD, Holm-Bonferroni
• Input: Numeric dependent variable + categorical grouping variable
• Outputs: ANOVA Summary · Pairwise Comparisons (Mean Difference, Corrected P-Value)

Heteroskedasticity Tests

Checks whether the residual variance of an OLS regression model is constant.

• Methods: Breusch-Pagan, White, Goldfeld-Quandt
• Input: Numeric target + predictor variables
• Outputs: Test Result · Model Summary · Data with Predictions and Residuals

Factorial ANOVA

Tests whether categorical factors — alone or in combination — significantly affect a continuous outcome.

• Features: Up to N-way interactions, Type I/II/III sums of squares, partial eta squared
• Input: Numeric response variable + one or more categorical factor variables
• Outputs: ANOVA Results (Basic or Advanced) · Model Coefficients with confidence intervals

One-Way MANOVA

Tests whether group means differ across multiple dependent variables simultaneously.

• Test statistic: Pillai's Trace (robust to assumption violations)
• Input: Two or more numeric dependent variables + one categorical grouping variable
• Outputs: Multivariate Results · Reliability Report (Box's M test)

Repeated Measures ANOVA

Tests whether the same participants respond differently across conditions or time points.

• Correction: Greenhouse-Geisser sphericity correction applied automatically
• Input format: Long format only — one row per measurement, with columns for the measured value, condition/time point, and participant ID
• Output: Basic summary or full Advanced breakdown (sphericity diagnostics included)

---

Instructions

Prerequisites

• KNIME Analytics Platform
• git
• pixi

Setup

1. Clone this repository:

   git clone <repository-url>
   cd knime-utd-statistics

2. Review knime.yml to confirm the extension metadata (name, group ID, author, version).

3. Inspect the src/ directory to explore or modify node implementations. Each node is implemented as a Python file (e.g. factorial_anova_node.py) backed by a dedicated submodule (e.g. factorial_anova/).

4. Install the Python environment:

   pixi install

   This installs all dependencies as defined in pixi.toml. The resulting environment is locked in pixi.lock — commit both files whenever you add or update packages.

5. (Optional) Add packages to the environment:

   pixi add <package_name>

6. Register the extension in debug mode with your local KNIME Analytics Platform:

   pixi run register-debug-in-knime

   This command auto-detects your KNIME installation and appends the -Dknime.python.extension.debug_knime_yaml_list argument to the knime.ini file automatically — no manual file editing required. You can then test the nodes in KNIME (see workflows/ for a demo).

7. Bundle your extension:

   pixi run build

   To place the update site at a custom path (default is ./local-update-site):

   pixi run build dest=<path_to_your_update_site>

8. Install the bundled extension in KNIME via:

   File > Install KNIME Extensions... > Available Software Sites > Add...
   

   Enter the path to your local update site. After that, install and restart KNIME.

9. To publish on KNIME Hub, follow the KNIME Hub documentation.

---

Testing

Golden reference tables and CSV fixtures live in golden_tables/, alongside scripts to regenerate them. Each node has a corresponding KNIME workflow in workflows/ (e.g. golden_NWay_ANOVA.knwf, golden_MANOVA.knwf) that can be used to validate output end-to-end inside KNIME Analytics Platform.

Run all tests with:

pixi run test

---

Join the Community

• KNIME Forum
• KNIME Python Extension Documentation