Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 2 results · all verified · 2 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs; all 2 also cleared it.

Fractional factorial aliases and foldover — Examples

1 · Prerequisites

2 · Summary

The first example computes all four alias classes in the 23−1 fraction with I=ABC. The second computes the exact residual relations after a complete mirror foldover and a one-factor follow-up. It contrasts resolution IV after the full mirror with a partial design that still aliases one main effect with an interaction.

3 · Logical flowchart

4 · Definitions, theorems and proofs

None yet.

5 · Examples, counterexamples and false statements

ExampleConstruction: AI-generatedVerification: AI-adaptedprecheck passjudge pass (gpt-6-luna)Open item page →

The alias classes in a 23−1 fraction

Example

Take k=3 with factors A,B,C and the positive generator equation xAxBxC=+1, conventionally written I=ABC. The four runs, in (A,B,C) order, are (−1,−1,+1), (−1,+1,−1), (+1,−1,−1), and (+1,+1,+1). The complete equal-column alias classes are

{I,ABC},{A,BC},{B,AC},{C,AB}.

In the unrestricted model with all eight factorial terms, each pair contributes through its coefficient sum. Separating a main effect from its aliased interaction therefore requires additional runs or an explicit sparsity or hierarchy assumption.

Facts & Assumptions

Given: k=3, the generator equation xAxBxC=+1, the four runs displayed in the Example, and the unrestricted fixed-design model with columns for all subsets of {A,B,C}.

[F1]

Regular two-level fractions, signed defining relations, and resolution: a regular fraction is specified by independent generator equations; H is generated by their signed words, and equal-column aliases are columns that agree on every run.

[F2]

The defining subgroup gives the exact alias relations: A and B have equal columns exactly when (+1,A△B)∈H.

[F3]

Aliased factorial columns prevent separate coefficient identification: if two distinct columns satisfy cA=σcB, their contribution depends only on σβA+βB, so the coefficients are not separately identifiable in the unrestricted model.

Verification

technique · enumerate the fraction and multiply its one defining word by every effect word
1.1givenF1constructchoose

Choose xA,xB∈{−1,+1} freely and set xC=xAxB. This gives exactly the four runs listed in the Example, one for each of the four choices of (xA,xB), and every run satisfies xAxBxC=+1.

2.1F1F2step 1.1algebra

Here H={I,ABC}. By [F2], effects have equal columns exactly when their symmetric difference is either ∅ or {A,B,C}. Pairing each subset with its complement gives precisely {I,ABC}, {A,BC}, {B,AC}, and {C,AB}; these four pairs exhaust the eight subsets.

3.1givenF3step 2.1algebra∎

In each listed pair the columns are equal, so [F3] shows that the corresponding terms enter the fitted vector through βI+βABC, βA+βBC, βB+βAC, or βC+βAB, respectively. The design by itself does not identify the two coefficients within any pair separately.

Sources

Oehlert §18.2 and NIST/SEMATECH §§5.3.3.4.1–5.3.3.4.3 give the standard I=ABC half-fraction and its alias notation. The four-row enumeration and coefficient ambiguity are verified locally.

ExampleConstruction: AI-generatedVerification: AI-adaptedprecheck passjudge pass (gpt-6-luna)Open item page →

Full and targeted foldovers of a resolution III fraction

Example

Use factors A,B,C,D,E and the regular 25−2 fraction generated by

xD=xAxB,xE=xCxD.

Writing I for the identity word, its defining relations are I=ABD and I=CDE, so H={I,ABD,CDE,ABCE} and the initial resolution is III. A complete mirror foldover yields the regular 25−1 fraction I=ABCE with H[5]={I,ABCE}: each subset S remains paired with S△{A,B,C,E}, so A remains aliased with BCE while C and DE separate. Reversing only factor A yields the regular 25−1 fraction I=CDE with H{A}={I,CDE}: each S remains paired with S△{C,D,E}, so A separates from BD but C remains aliased with DE. The targeted follow-up therefore leaves resolution III and a specific coefficient ambiguity; this is an exact residual alias list, not a claim that all main effects are clear.

Facts & Assumptions

Given: The five factors and generator equations in the Example, the complete mirror foldover J={A,B,C,D,E}, the partial foldover J={A}, and the unrestricted fixed-design linear model with all factorial columns on the augmented runs.

[F1]

Regular two-level fractions, signed defining relations, and resolution: signed words multiply by symmetric difference, and resolution is the shortest nonidentity defining-word length.

[F2]

The defining subgroup gives the exact alias relations: H is generated by the independent signed generators and gives all equal- or opposite-column alias relations.

[F3]

Full and partial foldovers separate aliases by parity: after reversing factor set J, the residual relation subgroup is HJ={(ϵ,S)∈H:∣S∩J∣ is even}.

[F4]

Aliased factorial columns prevent separate coefficient identification: equal columns in an unrestricted fixed-design model contribute only through the sum of their coefficients, and the two coefficients are not separately identifiable.

Verification

technique · multiply the generators, then apply the exact parity criterion to each follow-up
1.1givenF1F2algebra

The equations give xAxBxD=+1 and xCxDxE=+1, so the defining words are ABD and CDE. Their product is ABCE because the repeated D cancels in symmetric difference. The two incidence vectors are nonzero and distinct; over F2, two nonzero vectors are dependent only if equal, so these generators are independent. By [F2] these and the identity are exactly the four words in H. Their nonidentity lengths are 3,3,4, hence the initial resolution is III.

2.1givenF1F2F3step 1.1chooseconstructalgebra

For the complete mirror, [F3] changes the signs of the length-three relations ABD and CDE but preserves ABCE, giving H[5]={I,ABCE}. The two eight-run blocks are disjoint because ABD has opposite signs on them, and both lie in Q={x:xAxBxCxE=+1}; choosing xA,xB,xC,xD freely determines xE, so ∣Q∣=16 and their union is exactly this regular 25−1 fraction. Its aliases are precisely S with S△{A,B,C,E}, so A remains paired with BCE, while C△DE=CDE is absent and that pair separates. The shortest residual word has length 4.

2.2givenF1F2F3step 1.1chooseconstructalgebra

For the partial foldover on J={A}, [F3] changes ABD and ABCE but preserves CDE, giving HJ={I,CDE}. The two eight-run blocks are disjoint because ABD has opposite signs on them, and both lie in Q′={x:xCxDxE=+1}; choosing xA,xB,xC,xD freely determines xE, so their union is all 16 runs in this regular 25−1 fraction. The complete residual pairs are precisely S and S△{C,D,E}: this includes C=DE and excludes the original relation A=BD. The shortest residual word has length 3.

3.1givenF4step 2.2algebra∎

The retained columns cC=cDE in the partial design contribute as (βC+βDE)cC. By [F4], only this combination enters the fitted vector for this pair, so the partial follow-up does not separately identify βC and βDE without a restriction or further runs.

Sources

Oehlert §§18.2, 18.4, and 18.7 support the generator, resolution, and full-foldover conventions. NIST/SEMATECH §§5.3.3.8.1–5.3.3.8.2 covers mirror and selected-factor follow-ups. All relation products and residual alias sets in this example are computed explicitly.

Sources