Skip to content
Leantensify Learn

Which of these factors actually move the process — and do any of them depend on each other?

Run every combination of two-level factors and estimate every effect at once — including the interactions that changing one factor at a time is structurally unable to see.

Design of Experiments (2ᵏ Factorial) · improve · Green Belt

Use this when

  • Several candidate factors and no proof which ones matter
  • A suspicion that the right setting for one factor depends on another
  • Before a one-factor-at-a-time trial spends its runs finding a false optimum

Two-level factorial experiment

sandbox

5 effects distinguishable from noise — screening judgement, no error estimate exists

5 of 15 effects are distinguishable from noise (Temp, Conc, TempConc, Stir, TempStir), judged against Lenth's margin of 6.748 — a screening criterion, since an unreplicated design has no error estimate. The interactions TempConc, TempStir mean the factors involved cannot be set independently: the right setting for one depends on the other. The simple-effects table below shows each combination, and the recommendation has to be a PAIR of settings, not two separate ones. This is also exactly what one-factor-at-a-time experimentation cannot see. Before acting on this: run the winning combination again as a confirmation. An effect estimate is a prediction, and the confirmation run is where it meets the process.

Design
2^4, 1× replicated
Method
Lenth's method
Lenth PSE
2.625
Margin of error
6.748
Every effect this design can estimate, in standard order. An effect is the average change in the response between the low and high settings of its factors.
EffectEstimatevs margin 6.75Verdict
Temp21.625|21.625|Distinguishable
Pressure3.125|3.125|Noise-sized
TempPressure0.125|0.125|Noise-sized
Conc9.875|9.875|Distinguishable
TempConc-18.125|18.125|Distinguishable
PressureConc2.375|2.375|Noise-sized
TempPressureConc1.875|1.875|Noise-sized
Stir14.625|14.625|Distinguishable
TempStir16.625|16.625|Distinguishable
PressureStir-0.375|0.375|Noise-sized
TempPressureStir4.125|4.125|Noise-sized
ConcStir-1.125|1.125|Noise-sized
TempConcStir-1.625|1.625|Noise-sized
PressureConcStir-2.625|2.625|Noise-sized
TempPressureConcStir1.375|1.375|Noise-sized
Half-normal plot of 15 effect estimates. 5 effects peel away above the noise line: Conc at 9.88, Stir at 14.63, TempStir at 16.63, TempConc at 18.13, Temp at 21.63. The line's slope is the effect standard error, 2.625.noise lineConcStirTempStirTempConcTemp23.40Half-normal quantile|Effect|
Effects that estimate nothing fall on the dashed noise line; real ones peel away above it. Squares with labels are the effects the analysis calls distinguishable — the shape carries the meaning, not just the colour.

The interactions, decomposed — why one setting at a time is not a recommendation

Effect of Temp is 39.750 with Conc low, and 3.500 with Conc high.

Mean response at each combination of Temp and Conc.
Conc lowConc high
Temp low45.2573.25
Temp high85.0076.75

Effect of Temp is 5.000 with Stir low, and 38.250 with Stir high.

Mean response at each combination of Temp and Stir.
Stir lowStir high
Temp low60.2558.25
Temp high65.2596.50

This design is unreplicated, so there is no estimate of experimental error — every degree of freedom went into effects. Significance here comes from Lenth's method, which builds a noise floor from the smaller contrasts themselves. It is a screening judgement, good at finding the effects worth a confirmation run; it is not an F test, and a borderline effect deserves the replicated follow-up rather than a verdict.

Temp and Conc interact: the effect of Temp is 39.750 with Conc low and 3.500 with Conc high. The main effects of Temp and Conc average those two answers and describe neither; any recommendation has to name both settings together.

Temp and Stir interact: the effect of Temp is 5.000 with Stir low and 38.250 with Stir high. The main effects of Temp and Stir average those two answers and describe neither; any recommendation has to name both settings together.

Factor names
Run these in RANDOM order in real life; enter them here in this order.
#RunTempPressureConcStir
1(1)
2a+
3b+
4ab++
5c+
6ac++
7bc++
8abc+++
9d+
10ad++
11bd++
12abd+++
13cd++
14acd+++
15bcd+++
16abcd++++

Replicates on the same line, separated by commas.

Load a published experiment

How this is calculated
  • Effects are contrasts: the average response with the effect’s factors high minus the average with them low — effect = contrast / (n·2k−1), SS = contrast² / (n·2k). The contrasts are orthogonal, which is what lets 2k runs estimate 2k−1 effects independently.
  • Replicated designs test each effect with F on 1 and 2k(n−1) degrees of freedom against the pooled error.
  • Unreplicated designs have no error estimate — every degree of freedom went into effects. Significance uses Lenth’s pseudo standard error: PSE = 1.5 × median of the |effects| below 2.5 × (1.5 × median|effect|), with margin t(1−α/2, m/3) × PSE. It is a screening criterion, and the tool labels it as one rather than presenting it as an F test.
  • The half-normal plot ranks |effects| against half-normal quantiles. Effects that are pure noise fall on a line through the origin; real ones peel away above it.
  • Simple effects decompose each significant interaction, because the main effects it involves are averages of two different answers and cannot be quoted alone.

Source: Montgomery, D.C., Design and Analysis of Experiments 8e, Ch. 6 (Examples 6.1 and 6.2 are the presets, and the verification fixtures); Lenth, R.V. (1989), Technometrics 31(4), 469–473; Daniel, C. (1959) on the half-normal plot.

How this is calculated

Full 2^k factorial in standard (Yates) order. Each effect is an orthogonal contrast: effect = contrast/(n*2^(k-1)), SS = contrast^2/(n*2^k). Replicated designs test each effect by F on 1 and 2^k(n-1) degrees of freedom against the pooled error. Unreplicated designs have no error estimate — every degree of freedom went into effects — so significance uses Lenth's method: s0 = 1.5 x median|effect|, PSE = 1.5 x median of the |effects| below 2.5*s0, margin of error = t(1-alpha/2, m/3) x PSE, stated as a screening judgement rather than presented as an F test. The half-normal plot ranks |effects| against half-normal quantiles with the noise line drawn at the effect standard error. Every significant two-factor interaction is decomposed into simple effects with its cell means, because the main effects it involves are averages of two different answers and cannot be quoted alone.

Source: Montgomery, D.C., Design and Analysis of Experiments 8e, Ch. 6 — Examples 6.1 (plasma etch, replicated 2^3) and 6.2 (filtration, unreplicated 2^4) are the presets and the verification fixtures; Lenth, R.V. (1989), Quick and Easy Analysis of Unreplicated Factorials, Technometrics 31(4), 469-473; Daniel, C. (1959) on the half-normal plot.

Learn the method