senspy 0.2.0

Python port of the sensR package for Thurstonian models in sensory discrimination

sensPy

A Python port of the R package sensR for Thurstonian models in sensory discrimination analysis.

---

About

sensPy provides a comprehensive Python implementation of sensR, the R package for sensory discrimination methods. This project ensures:

• Numerical parity with sensR (validated against R outputs)
• Modern Python practices (type hints, dataclasses, NumPy/SciPy integration)
• Full test coverage with 500+ tests

Installation

# Install with pip
pip install senspy

# Or install from source
git clone https://github.com/aigorahub/sensPy.git
cd sensPy
pip install -e .

Features

Discrimination Protocols

All standard protocols with single and double variants:

Protocol	Single	Double
Triangle	✅	✅
Duo-Trio	✅	✅
2-AFC	✅	✅
3-AFC	✅	✅
Tetrad	✅	✅
Hexad	✅	-
2-out-of-5	✅	-

Statistical Models

• discrim() - D-prime estimation with confidence intervals and p-values
• betabin() - Beta-binomial model for overdispersed data
• twoac() - 2-Alternative Certainty model
• samediff() - Same-different model
• dod() - Degree-of-difference model
• anota() - A-not-A signal detection model

Power & Sample Size

• discrim_power(), dprime_power() - Power analysis
• discrim_sample_size(), dprime_sample_size() - Sample size calculation
• samediff_power() - Same-different power (simulation-based)
• twoac_power() - 2-AC exact power
• dod_power() - DOD power analysis

Inference & Comparison

• dprime_test() - Test d-prime hypotheses
• dprime_compare() - Compare d-primes across groups
• posthoc() - Post-hoc pairwise comparisons

ROC & Signal Detection

• roc() - ROC curve computation
• auc() - Area under the ROC curve
• sdt() - Signal detection theory transforms

Simulation

• discrim_sim() - Simulate discrimination experiments
• samediff_sim() - Simulate same-different data
• dod_sim() - Simulate DOD data

Visualization

• plot_psychometric() - Psychometric functions
• plot_roc() - ROC curves
• plot_power_curve() - Power curves
• plot_profile_likelihood() - Profile likelihood

Quick Start

from senspy import discrim, discrim_power

# Analyze a Triangle test result (80 correct out of 100)
result = discrim(correct=80, total=100, method="triangle")
print(f"d-prime: {result.d_prime:.3f}")
print(f"95% CI: [{result.ci_lower:.3f}, {result.ci_upper:.3f}]")
print(f"p-value: {result.p_value:.4g}")

# Calculate power for a future test
power = discrim_power(d_prime=1.5, sample_size=100, method="triangle")
print(f"Power: {power:.1%}")

from senspy import samediff, dprime_compare

# Same-different analysis
sd = samediff(nsamesame=45, ndiffsame=5, nsamediff=20, ndiffdiff=30)
print(f"delta: {sd.delta:.3f}, tau: {sd.tau:.3f}")

# Compare d-primes across products
result = dprime_compare(
    correct=[80, 65, 90],
    total=[100, 100, 100],
    method="triangle"
)
print(f"Chi-square: {result.statistic:.2f}, p={result.p_value:.4f}")

Disclaimer

Beta Release — sensPy is currently in beta. While we strive for numerical accuracy and validate results against the original sensR package, this software is provided "as is", without warranty of any kind. Aigora assumes no liability for any decisions, outcomes, or consequences arising from the use of this software. Users are responsible for independently verifying results before relying on them in production, regulatory, or research contexts.

License

This project is licensed under the GNU General Public License version 2.0 or later. See LICENSE for details.

Citation

If you use sensPy in your research, please cite:

@software{senspy,
  title = {sensPy: Python port of sensR for Thurstonian models},
  author = {Aigora},
  year = {2025},
  url = {https://github.com/aigorahub/sensPy}
}

Acknowledgments

This package is a port of sensR by Per Bruun Brockhoff and Rune Haubo Bojesen Christensen.