Minitab Brand-Support PageINDEPENDENT STATISTICAL REVIEW

Minitab Factorial Design: An Independent DOE Guide

Explore how factorial designs in Minitab Statistical Software support the study of multiple factors and interaction effects.

An independent educational review of experiment planning, response data, and interpretation.

Methodology & Practice

Foundations of Designed Experiments in Minitab

An independent review of factorial planning structures, level assignments, and the anatomy of screening runs.

What a Designed Experiment Is

A Designed Experiment (DOE) is a planned sequence of runs in which several input variables are changed deliberately and at the same time, so that their individual and joint influence on an outcome can be separated from background noise.

Core Principle

Factors and Responses

Factors are the inputs the experimenter controls (such as a temperature setpoint or a feed rate). The response is the measured outcome recorded for every run. In a Minitab worksheet, each factor and each response occupies its own column.

Variables

Factor Levels

Each factor is tested at chosen settings called levels. Two-level screening designs usually code them as low (-1) and high (+1), which keeps the design matrix balanced and makes effects easy to compare across factors measured in different units.

Levels (-1, +1)

Full Factorial Designs

A full factorial design runs every combination of factor levels. With k factors at two levels this means 2k runs, so all main effects and all interactions can be estimated without aliasing.

2k Runs

Fractional Designs

When many factors are involved, a fractional factorial (2k-p) runs a balanced subset of the combinations. It saves runs, but some effects become aliased with each other; the design resolution describes which effects can still be separated.

Screening

Randomization and Replication

Randomizing the run order protects results from time-related drift such as warm-up or tool wear. Replicates — complete, independent repeats of a run — give an estimate of pure error, which is needed to judge whether an effect is larger than ordinary variation.

Statistical Rigor

Main Effects

A main effect is the change in the average response when one factor moves from its low level to its high level, averaged over the levels of the other factors. A main effects plot in Minitab draws these averages as connected points for each factor.

  • A steeper line means a larger average change in the response.
  • Main effects can mislead when a strong interaction is present.

Interaction Plots

An interaction plot shows the mean response for one factor at each level of another. Non-parallel lines suggest that the effect of Factor A depends on the setting of Factor B.

  • Roughly parallel lines suggest little or no interaction.
  • Clearly non-parallel or crossing lines suggest an interaction worth testing.
  • The plot is descriptive; significance comes from the fitted model.

Analysis Assumptions

Conclusions from a factorial model rely on assumptions about the residuals. Minitab provides residual plots to review them before interpreting effects:

  • Normality: residuals fall close to a straight line on the normal probability plot.
  • Constant variance: residuals versus fits show no funnel pattern.
  • Independence: residuals versus run order show no drift or clustering.
Illustrative Experimental Matrix

2³ Full Factorial Design Layout

Filter the combinations below to inspect the coded treatment structure of an eight-run, three-factor design.

Swipe to view all columns

Run #Standard OrderTemperature (°C)Pressure (bar)Feed RateSynthetic Output (Units)*
Run 01Std 1-1 (140°C)-1 (2.0 bar)-1 (Low)41.8
Run 02Std 2+1 (180°C)-1 (2.0 bar)-1 (Low)54.2
Run 03Std 3-1 (140°C)+1 (3.5 bar)-1 (Low)46.5
Run 04Std 4+1 (180°C)+1 (3.5 bar)-1 (Low)67.9
Run 05Std 5-1 (140°C)-1 (2.0 bar)+1 (High)49.1
Run 06Std 6+1 (180°C)-1 (2.0 bar)+1 (High)61.7
Run 07Std 7-1 (140°C)+1 (3.5 bar)+1 (High)52.3
Run 08Std 8+1 (180°C)+1 (3.5 bar)+1 (High)78.4

*Data origin: the output values in this table are a synthetic teaching dataset written by the FactorInteraction Notebook editorial team to illustrate the layout of a 2³ worksheet. They are not measurements from a real process and not results produced by Minitab Statistical Software, and no p-values or conclusions are drawn from them.

View Case Studies
I. Brand-Specific Scenario

A Two-Factor Experimental Matrix

How teaching factors and their level combinations can be organized as a factorial design in Minitab Statistical Software — before any response data is collected.

Planning the Design

In this teaching scenario two factors are studied at two levels each, giving a 2² full factorial with four distinct treatment combinations:

  • Factor A — Temperature: low 140 °C (-1), high 180 °C (+1).
  • Factor B — Pressure: low 2.0 bar (-1), high 3.5 bar (+1).
  • Response: process yield, recorded by the experimenter for every run.

A typical path in Minitab is Stat > DOE > Factorial > Create Factorial Design: choose a 2-level factorial with 2 factors, select the full factorial, set the number of replicates, and enter the factor names and level values. Minitab then generates a worksheet with standard order, run order and the factor columns; the response column is added and filled in after the runs are performed.

Menu names may differ between Minitab releases; consult the official Minitab documentation for your version.

Design Worksheet Structure

2² Full Factorial, 2 Replicates (8 Runs)

Run order shown is one example of a randomized sequence; Minitab creates its own random order each time a design is generated.

Swipe to view all columns

Std OrderRun OrderA: TemperatureB: PressureYield
15-1 (140°C)-1 (2.0 bar)to be measured
22+1 (180°C)-1 (2.0 bar)to be measured
38-1 (140°C)+1 (3.5 bar)to be measured
41+1 (180°C)+1 (3.5 bar)to be measured
57-1 (140°C)-1 (2.0 bar)to be measured
63+1 (180°C)-1 (2.0 bar)to be measured
76-1 (140°C)+1 (3.5 bar)to be measured
84+1 (180°C)+1 (3.5 bar)to be measured

What the matrix allows: with both factors at both levels and two replicates, the analysis can estimate the main effect of A, the main effect of B, the A×B interaction, and pure error from the replicates.

No results are reported here. This scenario shows only the structure of the design. Effects, p-values and conclusions can only come from analyzing real measured responses in Minitab (Stat > DOE > Factorial > Analyze Factorial Design); this guide does not substitute invented results for them.