PyDOE: An Experimental Design Package for Python

0.9.9 · active · verified Sat Apr 11

PyDOE is a Python package for design of experiments (DOE), enabling scientists, engineers, and statisticians to efficiently construct experimental designs. It provides extensive support for various DOE methods, including factorial, response-surface, and space-filling designs. The project is actively maintained, with a focus on integrating features from its community forks into the main package.

Warnings

breaking Older versions of `pydoe` (prior to 0.9.x) were not compatible with Python 3 directly, primarily due to changes in the division operator. Attempts to use them with Python 3.x often resulted in `SyntaxError` or `TypeError`. The current version (>=0.9.x) explicitly requires Python >=3.10.
Fix: Ensure you are using `pydoe` version 0.9.0 or higher with Python 3.10+ (`pip install pydoe`). If maintaining older Python 3 versions, consider `pyDOE2` or manually patching older `pydoe` versions (though not recommended).
gotcha Historically, several forks like `pyDOE2` and `pyDOE3` emerged due to perceived inactivity or unaddressed bugs in the original `pydoe`. While these forks offered updates and new features, the official `pydoe` project is now actively maintained and aims to integrate these improvements.
Fix: For new projects, prefer the official `pydoe` package (`pip install pydoe`). Existing users of `pyDOE2`/`pyDOE3` are encouraged to monitor `pydoe`'s development for migration back to the main branch for long-term support.
gotcha In `pydoe` versions before 0.9.x, users might have encountered `DEPRECATION: pyDOE is being installed using the legacy 'setup.py install'` warnings due to the absence of a `pyproject.toml` file.
Fix: This issue has been addressed in `pydoe` version 0.9.x and newer, which includes `pyproject.toml` for modern build processes. Upgrade to the latest `pydoe` version to resolve this (`pip install --upgrade pydoe`).
gotcha Some functions, like `ccdesign`, in older `pydoe` versions were reported to have issues with specific `numpy` versions, leading to `TypeError: can't multiply sequence by non-int of type 'numpy.float64'`.
Fix: Ensure both `pydoe` and its dependencies (`numpy`, `scipy`) are updated to their latest compatible versions. If an issue persists, consult the `pydoe` GitHub issues or documentation for known compatibility notes.

Install

pip install pydoe Install latest version

Imports

ff2n
```
from pydoe import ff2n
```
For 2-level full-factorial designs.
lhs
```
from pydoe import lhs
```
For Latin Hypercube Sampling.
ccdesign
```
from pydoe import ccdesign
```
For Central-Composite Designs.

Quickstart

This quickstart demonstrates how to generate a 2-level full-factorial design and Latin Hypercube samples using pydoe. The `ff2n(n)` function creates a 2^n factorial design, and `lhs(n, samples)` generates 'samples' points for 'n' variables using Latin Hypercube Sampling.

import numpy as np
from pydoe import ff2n

# Create a 2^3 full-factorial design (3 factors, each at 2 levels)
design = ff2n(3)
print("2^3 Full-Factorial Design Matrix:")
print(design)

# Example of Latin Hypercube Sampling (LHS) for 2 variables, 5 samples
from pydoe import lhs
lhs_samples = lhs(2, samples=5)
print("\nLatin Hypercube Samples (2 variables, 5 samples):")
print(lhs_samples)

view raw JSON →