cpistats 0.1.1

Statistical tools for Continuous Process Improvement and Lean Six Sigma

cpistats

Statistical tools for Continuous Process Improvement (Lean Six Sigma)

  

---

Overview

cpistats is a Python package that provides statistically rigorous and reproducible methods commonly used in Continuous Process Improvement (CPI) and Lean Six Sigma projects.

The library is built around three core principles:

• Correct statistical implementation — methods aligned with established references
• Reproducibility — deterministic, fully parameterized outputs
• Validation — results verified against NIST and Minitab benchmarks

---

Installation

pip install cpistats

---

Quick Start

from cpistats import two\\\_sample\\\_t\\\_test

data1 = \\\[1.4, 1.3, 1.5, 1.3]
data2 = \\\[1.3, 1.2, 1.4, 1.2]

result = two\\\_sample\\\_t\\\_test(
    data1,
    data2,
    delta0=0.1,
    alternative="greater",
    equal\\\_var=True,
)

print(result.t\\\_statistic)
print(result.p\\\_value)

---

Implemented Methods

Hypothesis Tests

Method	Description
anderson\\\_darling\\\_normality	Anderson–Darling normality test
one\\\_sample\\\_t\\\_test	One-sample t-test
two\\\_sample\\\_t\\\_test	Two-sample t-test (pooled or Welch)

Two-sample t-test variants:

• Pooled — assumes equal variances (equal\\\_var=True)
• Welch — assumes unequal variances (equal\\\_var=False)

---

Validation

All methods are validated against trusted industry references:

• NIST Statistical Reference Datasets (StRD) — numerical accuracy benchmarks
• Minitab (v22) — industry-standard statistical software

Validation Framework

Each validated dataset includes:

• Input data and parameters
• Reference results from the trusted source
• Computed Log Relative Error (LRE) and Log Absolute Error (LAE)

Automated validation is run on every change to ensure numerical accuracy and consistency are maintained.

---

Project Structure

src/cpistats/
├── hypothesis\\\_tests/
│   ├── anderson\\\_darling\\\_normality.py
│   ├── one\\\_sample\\\_t\\\_test.py
│   └── two\\\_sample\\\_t\\\_test.py
├── core/
└── validation/

---

Design Principles

• Deterministic outputs — same inputs always produce the same results
• Explicit parameter control — no hidden defaults (e.g., delta0, alternative, equal\\\_var)
• Industrial alignment — behavior matches Minitab and other trusted tools
• Transparent and testable logic — clean, auditable code

---

Roadmap

Planned additions:

• [ ] Paired t-test
• [ ] Nonparametric tests (Sign test, Wilcoxon signed-rank test)
• [ ] Confidence intervals
• [ ] Effect size metrics (Cohen's d, Hedges' g)
• [ ] Statistical Process Control (SPC) charts

---

License

This project is licensed under the MIT License.

---

Author

Alexandre Torres