The Four-Run Factorial Example
A complete step-by-step examination of the classic 2-squared matrix, demonstrating balance properties, sign vector multiplication, and interaction calculation.

Contents Overview
The 2×2 Design Matrix
The smallest factorial experiment has two factors, A and B, each at two levels. The four combinations are written in standard order below. The AB column is the product of the A and B signs.
Each column contains two minus signs and two plus signs, and any two columns multiplied row by row sum to zero. This balance, called orthogonality, is what allows the effects to be estimated independently.
Swipe to view all columns
| Run (std order) | A | B | AB | Response |
|---|---|---|---|---|
| 1 | −1 | −1 | +1 | y1 |
| 2 | +1 | −1 | −1 | y2 |
| 3 | −1 | +1 | −1 | y3 |
| 4 | +1 | +1 | +1 | y4 |
Key Principle
The effect of one factor may depend on the level of another.
Calculating Effects Symbolically
Call the responses of the four runs y1, y2, y3 and y4. The main effect of A is the average response at high A minus the average at low A: (y2 + y4)/2 − (y1 + y3)/2. The main effect of B is (y3 + y4)/2 − (y1 + y2)/2.
The interaction AB uses the AB column: (y1 + y4)/2 − (y2 + y3)/2. If the effect of A is the same at both levels of B, this quantity is close to zero.
What Four Runs Cannot Tell You
With four runs and four estimated quantities (mean, A, B, AB) there are no degrees of freedom left to estimate error. Without replication there is no basis for judging whether an effect is larger than ordinary variation.
The four-run example is therefore a tool for understanding structure. For real decisions, replicate the runs and randomize their order.
Key Takeaways
- Standard order lists combinations; it is not a prescribed run order.
- Orthogonal columns allow independent effect estimates.
- Unreplicated four-run designs leave no estimate of error.