cbm_power

Computational Behavioral/Brain Modeling (CBM) Library for power analysis and sample-size optimization for computational studies.

Overview

cbm_power is a Python package for estimating statistical power and optimizing sample sizes in computational modeling studies.
It provides tools for simulation-based power estimation and automated sample-size optimization. It implements the method described in this study: Piray Payam, Addressing low statistical power in computational modeling studies in psychology and neuroscience, 2025, Nature Human Behaviour.

Installation

Clone this repository and install dependencies:

git clone https://github.com/your-username/cbm_power.git
cd cbm_power
pip install -e .

Dependencies include:

• numpy
• ax-platform for Bayesian optimization

Usage

from cbm_power import Power, SampleSize, Config

# Run power estimation
pwr = Power()
power, _ = pwr.compute_power(
    num_participants=410,
    num_models=10
)

print("Estimated power:", power)

# Optimize sample size
cfg = Config(num_models=10)  # Create a config first
ss = SampleSize(cfg)
best_sample_size = ss.compute_sample_size()  # this saves a JSON and a log file
print("Optimized sample size:", best_sample_size)

Examples

• examples/example_power.py — run power analysis for a given sample size and model space size.
• examples/example_sample_size.py — optimize N and save JSON/log artifacts.

Repository structure

cbm_power/
├── cbm_power/
│   ├── __init__.py        # Exposes Power, SampleSize, Config
│   ├── config.py          # Config dataclass used in SampleSize
│   ├── distributions.py   # Distribution utilities
│   ├── logger.py          # Logging utilities
│   ├── optimizer.py       # Internal optimizer helpers
│   ├── power.py           # Power estimation logic
│   └── sample_size.py     # Sample size optimization
├── docs/
│   └── manual.md          # Markdown version of the manual
├── examples/
│   ├── example_power.py           # Power computation example
│   └── example_sample_size.py     # Sample-size optimization ├── README.md
├── LICENSE
├── pyproject.toml
└── requirements.txt

Output

Running sample-size optimization automatically saves:

• JSON file with results
• Log file with optimization trace

Both are stored in the current directory.

Documentation

A manual explaining the method, background, and step-by-step usage of cbm_power is available in the docs/ directory.
It includes:

• Key conceptual points about power in computational modeling studies
• Explanation of the multi-stage optimization method
• Worked examples with the Power and SampleSize classes
• Troubleshooting guidance (performance, memory, and numerical stability)

License

MIT License.

Reference

If you use this package in your research, please cite the following paper:

Piray, Payam, "Addressing low statistical power in computational modeling studies in psychology and neuroscience", Nature Human Behaviour, 2025.