Comparison of different definitions of regular fraction
Maria Piera Rogantin
DIMA – Universit`a di Genova – [email protected]
Giovanni Pistone
DIMAT – Politecnico di Torino – [email protected] SCRA 2006, Tomar - Portugal, September 1-4, 2006
Regular fractions in literature
• Generating words:
– If the level number is a prime number, n = p: generating words in
“additive” notation and computation in Zp (integer coding)
– If the level number is a power of a prime number, n = ps: gener- ating words and computation in GF(ps) [e.g.: Dey and Mukerjee (1999)]
– Complex coding: generating words in “multiplicative” notation and computation in C [Bailey (1982), . . . , Kobilinsky (1997), . . . , Pistone and Rogantin (2005)]
• Property of non-existence of partial aliasing (associated to a spe- cific basis of responses) [e.g.: Wu and Hamada (2000)]
Assumption:
symmetrical fractional factorial designs (same number of levels n)
Regular fractions in complex coding
Pistone, G., Rogantin, M.-P., 2005. Indicator function and complex cod- ing for mixed fractional factorial designs. Submitted to JSPI.
Complex coding for levels
We code the n levels of a factor A with the n-th roots of the unity:
ωk = exp
i 2π n k
for k = 0, . . . , n − 1 Ωn = {ω0, . . . , ωn−1}
w
w w1
0
2
Regular fraction
:two factors or interactions are either orthogonal or totally aliased
Outline
• All definitions are equivalent if all factors have the same prime num- ber of levels n = p
• Case n = ps:
G
F
S ( Ω
ps)
mC-regular fractions in
( Ω
p)
smC-regular fractions
in [3 levels - 4 factors]
(GF[
ps])
mGF-regular fractions in [9 levels - 2 factors]
[9 levels - 2 factors]
Notations
• D: the full factorial design with complex coding D = Ωmn
L: the full factorial design with integer coding L = Zmn
L is also the exponent set of all the interactions α, β, . . .: the elements of L:
L = nα = (α1, . . . , αm) : αj = 0, . . . , n − 1, j = 1, . . . , mo
• Xi: a factor
Xα = X1α1 · · · Xmαm interactions or simple factors
Full factorial design and fractions
The full factorial design D, as a subset of Cm, is defined by the system of equations
Xjn − 1 = 0 for j = 1, . . . , m
A fraction is a subset of a full factorial design, F ⊂ D.
It is obtained by adding equations (generating equations) to restrict the set of solutions.
Indicator function F of a fraction F
It is a response defined on the full factorial design D such that F (ζ) =
1 if ζ ∈ F
0 if ζ ∈ D r F It is represented as polynomial as:
F = X
α∈L
bαXα(ζ) ζ ∈ D
Important statistical features of the fraction can be read out from the form of the polynomial representation of the indicator function.
C -regular fractions
• F a fraction without replicates where all factors have n levels
• Ωn the set of the n-th roots of the unity, Ωn = {ω0, . . . , ωn−1}
• L a subset of exponents, L ⊂ L = (Zn)m containing (0, . . . , 0), l = #L
• e a map from L to Ωn, e : L → Ωn
A fraction F is regular if:
1. L is a sub-group of L,
2. e is a homomorphism, e([α + β]) = e(α) e(β) for each α, β ∈ L,
3. the defining equations are of the form
Xα = e(α) , α ∈ L
If H is a minimal generator of the group L, then the equations Xα = e(α) with α ∈ H ⊂ L are called minimal generating equations.
• We consider the general case where e(α) can be different from 1
• We have no restriction on the number of levels
A necessary condition is the e(α)’s must belong to the subgroup spanned by the values Xα.
For example for n = 6 an equation like X13X23 = ω2 can not be a defining equation.
w2
w0 w1
w3
w5
w4
w2
w0 w1
w3
w
5
w
4
Subgroup spanned by: Xi2 Xi3
Example 1: A regular fraction of a 34 design
A B C D
0 0 0 0
0 1 1 1
0 2 2 2
1 0 1 2
1 1 2 1
1 2 0 1
2 0 2 1
2 1 0 2
2 2 1 0
X1 X2 X3 X4
1 1 1 1
1 ω1 ω1 ω1
1 ω2 ω2 ω2
ω1 1 ω1 ω2
ω1 ω1 ω2 1 ω1 ω2 1 ω1
ω2 1 ω2 ω1
ω2 ω1 1 ω2
ω2 ω2 ω1 1
The complex fraction is defined by
Xj3 − 1 = 0 , j = 1, . . . , 4 together with the generating equations
X1X2X32 = 1 and X1X22X4 = 1 . Then: H = {(1, 1, 2, 0), (1, 2, 0, 1)}
e(1, 1, 2, 0) = e(1, 2, 0, 1) = ω0 = 1
L = {(0, 0, 0, 0), (0, 1, 1, 2), (0, 2, 2, 1), (1, 1, 2, 0),
(2, 2, 1, 0), (1, 2, 0, 1), (2, 1, 0, 2), (1, 0, 1, 1), (2, 0, 2, 2)}.
Indicator function and regular fractions
(Pistone, Rogantin, 2005)
The following statements are equivalent:
1. The fraction F is regular according to previous definition with defining equations Xα = e(α), α ∈ L
2. The indicator function of the fraction has the form F (ζ) = 1
l
X α∈L
e(α) Xα(ζ) ζ ∈ D
where L is a given subset of L and e : L → Ωn is a given mapping.
3. For each α, β ∈ L the parametric functions represented on F by the terms Xα and Xβ are either orthogonal or totally aliased
Example 1 (continued): A regular fraction of a 34 design
The indicator function is:
F = 1
9
1 + X2X3X4 + X22X32X42 + X1X2X32 + X12X22X3 +X1X22X4 + X12X2X42 + X1X3X42 + X12X32X4
Example 2: a regular fraction of a 63 design The terms Xα take values in Ω6 or one of the two subgroups {1, ω3} and {1, ω2, ω4}.
The generating equations of the fraction are X13X33 = ω3 and X24X24X32 = ω2
w2
w0
w1
w3
w
5
w
4
(subgroups – laterals)
Then: H = {(3, 0, 3), (4, 4, 2)}
e(3, 0, 3) = ω3, e(4, 2, 2) = ω2
L = {(0, 0, 0), (3, 0, 3), (4, 4, 2), (2, 4, 4), (1, 4, 5), (5, 2, 1)}.
The full factorial design has 216 points and the fraction has 36 points The indicator function is:
F = 1
6 1 + ω3X13X33 + ω4X14X24X32 + ω2X12X22X34 + ω1X1X24X35 + ω5X15X22X3
C -regularity under permutation of levels.
A C-regular fraction is transformed into a C-regular fraction when a level permutation of the following type is performed on one or more factors:
1. Cyclical permutations on the factor Xj:
(ζ1, . . . , ζj, . . . , ζm) 7→ (ζ1, . . . ,ωkζj, . . . , ζm) k = 0, . . . , nj − 1 All the n level cyclical permutations are obtained in this way.
2. If nj is a prime number, permutations on the factor Xj: (ζ1, . . . , ζj, . . . , ζm) 7→ (ζ1, . . . , ωkζjh, . . . , ζm) . with k = 0, . . . , nj − 1 and h = 1, . . . , nj − 1.
In this way a sub-group of permutation of order nj(nj− 1) is obtained.
If n = 2 or n = 3 all the level permutations are of type 2.
Example 1 (continued): A regular fraction of a 34 design
We permute the levels ω0 and ω1 of the last factor X4 using the trans- formation:
(ζ1, ζ2, ζ3,ζ4) 7→ (ζ1, ζ2, ζ3, ω1ζ42) .
The indicator function of the transformed fraction is:
F = 1
9
1 + ω1X2X3X42 + ω2X22X32X4 + X1X2X32 + X12X22X3 +ω1X1X22X42 + ω2X12X2X4 + ω2X1X3X4 + ω1X12X32X42 . Then it is a regular fraction.
The generating equations of the transformed fraction are:
X1X2X32 = 1 and X1X22X42 = ω2 .
Regular fractions in Galois Field theory
for p
slevels
Let G be a fraction generated by the following equations in [GF (ps)]m:
m X j=1
aijzj = bi i = 1, . . . , r aij, zj, bi ∈ GF (ps)
g irreducible polynomial such that GF (ps) is isomorphic to the field Zp[x]/ < g(x) >
Example: GF(22)
The Galois Field GF(22) can be represented as the field of equivalence classes mod x2 + x + 1 of polynomials of degree 1 whose coefficients belong to Z2:
GF(22) ∋ z −→ z0 + z1x z0, z1 ∈ Z2 Polynomial representation of GF(22): {0, 1, x, 1 + x}.
GF-regular and C -regular fractions for p s levels
Recoding of each factor:
GF (ps) ∋ z = z0 + z1x + · · · + zs−1xs−1 −→ (ωz0, ωz1, . . . , ωzs−1) ∈ (Ωp)s The GF-regular fraction G is mapped onto a C-regular fraction F
F has s × m factors: each original factor splits into s pseudo-factors
G F
GF-regular fractions in
[4 levels - 2 factors]
[9 levels - 2 factors]
(GF[ p ])s m
C-regular fractions in
[2 levels - 4 factors]
[3 levels - 4 factors]
Ω
p( )
smNot all C-regular fractions in the pseudo-factors correspond to GF-regular fractions. Special conditions have to be assumed.
Example: GF-regular fractions with n = 22
Regular fractions in GF(22) generated by:
a1z1 + a2z2 = 0 a1, a2, z1, z2 ∈ GF(22)
(a10 + a11x)(z10 + z11x) + (a20 + a21x)(z20 + z21x) = 0 mod (x2 + x + 1) aij ∈ Z2 Generating equations in the pseudo-factors with levels in Ω2:
X10a10X11a11X20a20X21a21 = 1 X10a11X11a10+a11X20a21X21a20+a21 = 1
There are two inequivalent fractions:
G1 :
Z1 Z2
1 + x 1 + x
1 1
x x
0 0
F1 :
X10 X11 X20 X21
−1 −1 −1 −1
−1 1 −1 1
1 −1 1 −1
1 1 1 1
F1 = 1 4 + 1
4X10X20 + 1
4X11X21+ 1
4X10X11X21X20
G2 :
Z1 Z2
1 + x x 1 1 + x
x 1
0 0
F2 :
X10 X11 X20 X21
−1 −1 1 −1
−1 1 −1 −1 1 −1 −1 1
1 1 1 1
F2 = 1 4 + 1
4X10X21 + 1
4X11X20X21+ 1
4X10X11X20
Pseudo factors for C -regular fractions
A regular fraction S with m factors and ps levels can be transformed in a regular fraction F with s × m factors where each original factor splits into s pseudo-factors. The generating equations are transformed by the map
Ωps ∋ Yj = Xj0Xj1(1/p)Xj2(1/p)2 · · · Xjs−1(1/p)(s−1) ∈ (Ωp)s
C-regular fractions in
[4 levels - 2 factors]
[9 levels - 2 factors]
Ω
p( )
mC-regular fractions in
[2 levels - 4 factors]
[3 levels - 4 factors]
Ω
p( )
sm sF S
Example:
C-regular fractions in [Ω4]2 splitting in C-regular fractions in [Ω2]4 with generating equations of the form
Y α = 1
all C-r.f. in [Ω4]2 special C-r.f. in [Ω2]4 [GF (4)]2 S1 : Y1Y2 = 1 X10X20 = 1 X10X11X21 = 1
S2 : Y1Y22 = 1 X10X20 = 1 X11X22 = 1
S3 : Y1Y23 = 1 F1: X10X20 = 1 X11X22 = 1 G1 S4 : Y12Y22 = 1 X10X20 = 1
F2: X10X21 = 1 X11X20X21 = 1 G2 Only S3 corresponds to a GF–regular fraction, G1
The GF–regular fraction G2 does not correspond to any Si
Discussion
1. The definition of C-regular fractions applies to asymmetric case.
2. A fraction is C-regular if and only if the complex interactions have the non partial aliasing property.
3. Proposition about C-regular fractions provides an algorithmic way to check if a given array is a regular fraction.
4. If the number of levels is:
•
p
prime: Zp-regular fractions are C-regular fractions and vicev- ersa.•