Understanding Experimental Factor Levels

Learn how to select appropriate low and high setpoints, avoid nonlinear regions, and maintain dimensional stability across multi-variable experimental spaces.

2026-09-12
Marcus Hale
4 min read
Parameter Setup4 min read
Published: 2026-09-12
Understanding Experimental Factor Levels

Contents Overview

Article Sections

What a Level Is

A level is a specific setting of a factor used in the experiment. A temperature factor might be tested at 140 °C and 180 °C; a categorical factor such as surface preparation might be tested as standard and modified.

In two-level designs the settings are usually coded as -1 for low and +1 for high. Coding makes factors with different units comparable and keeps the design matrix balanced.

Key Principle

The effect of one factor may depend on the level of another.


How Far Apart Should Levels Be?

Levels that are too close together may produce a change in the response that is smaller than ordinary measurement variation, so a real effect looks like noise. Levels that are too far apart may move the process into unsafe, unstable or strongly nonlinear regions.

A good starting point is the widest range that is still practical, safe and relevant to how the process will actually be run.


Two Levels and the Question of Curvature

Two levels can only describe a straight-line change between low and high. If the true relationship curves, a two-level design will not show it. Adding center points, runs at the midpoint of all continuous factors, gives a simple check for curvature without changing the factorial structure.

Record the reasoning behind each level choice. Reviewers should be able to see why a range was chosen, not just what it was.


Key Takeaways

  • Levels are the specific settings tested for each factor.
  • Spacing must be wide enough to see an effect and narrow enough to stay practical.
  • Center points help check for curvature in two-level designs.
Marcus Hale

Marcus Hale

Contributor

Educational article for FactorInteraction Notebook. It explains planning logic and does not report experimental results or statistical significance.

Editorial Method