Factors, levels, responses, interactions, and noise

Learning objectives

• Explain the purpose of factors, levels, responses, interactions, and noise in a garment-factory improvement project.
• Apply the described method to a representative shop-floor situation.
• Recognize the common mistakes and how to avoid them.

Concept

Clear factor ranges and reliable responses determine whether an experiment can answer the factory question.

Factor = input X; level = tested setting; response = Y; interaction = one factor's effect depends on another; noise = uncontrolled variation.

Levels too narrow hide real effects; too wide risk unsafe or unrepresentative operation.

Variables, units, and assumptions

Factor X in engineering units; level as coded ±1 or actual value; response Y in measurable units.

Levels representative of production; measurement gauge verified.

Free-spreadsheet workflow

Tabulate factor/level/response; compute mean Y at each combination; plot interaction (mean Y vs factor A, one line per level of B).

Interpretation and limitations

Non-parallel interaction lines indicate interaction; parallel lines indicate main effects only.

Interaction estimation requires the design to include the relevant combinations.

Garment-factory example

Seam strength may respond differently to SPI at different thread tickets — an interaction missed by one-factor-at-a-time trials.

Method

1. Use process knowledge and risk review.
2. Set feasible level ranges.
3. Operationally define Y and measurement.
4. Record nuisance variables.
5. State interaction hypotheses in advance.

Common mistakes

• Choosing levels too narrow to reveal effects.
• Selecting an easy but irrelevant response.