STAT 512 2-Level Factorial Experiments: Irregular Fractions IRREGULAR TWO-LEVEL FRACTIONAL FACTORIALS • A major practical weakness of “regular” fractional factorial designs is that N must be a power of 2: 8 16 32 64 128 (large gaps) • A broader class of 2-level designs for which low-order effects are orthogonally estimable when higher-order effects are assumed zero: 1 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Orthogonal Array of strength t, OA(t): Design for which every subset of t factors (ignoring the rest) forms a full 2t factorial design, possibly replicated – less rigid structure requirement than for regular ff’s – we don’t require that every pair of higher-order effects be either orthogonal or completely confounded – allows for more flexibility in the value of N , e.g. an OA(2) could be of size 12 (every pair of factors a 22 × 3 reps ...) – generally not so easy to construct as regular fractional factorials • A useful class of OA’s of strength 2: 2 STAT 512 2-Level Factorial Experiments: Irregular Fractions 3 PLACKETT-BURMAN DESIGNS (Biometrika 1946) • Resolution III designs (estimable main effects in the absence of interactions) = 0 mod[4] • Available for N ≥ f +1 • Characterized by first row of design matrix: N first row 8 + + + − + − − 12 + + − + + + − 16 . . . (through N ≈ 100 in the PB paper) − − + − STAT 512 2-Level Factorial Experiments: Irregular Fractions • Construction of N -run design matrix D: 1. use given row for first run 2. use cyclic permutations of the given row for runs 2 through N − 1 3. use (− − −...−) for N th run 4. can (and usually should) randomize: – columns in the resulting design matrix to the physical experimental factors. – order of run execution 4 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Example, N = 8, f = 7: M = (1, D) = + + + + − + + + − − + − − + − + + − − + + + − + + + − − + + + − + − + − − + + + + + − + − − + + + + + − + − − + + − − − 5 − − − − STAT 512 2-Level Factorial Experiments: Irregular Fractions • For f = 4, 5, 6, use a subset of these columns −s • For N = 2c , these are equivalent to 2fIII regular fractions • For N 6= 2c , these are nongeometric, still providing orthogonal estimates of main effects, but alias structure is more complex 6 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Review of Bias in Linear Models Suppose y = X1 θ 1 + X2 θ 2 + , usual assumptions ... But we fit parameters as though model is y = X1 θ 1 + θ̂ 1 = (X01 X1 )−1 X01 y E[θ̂ 1 ] = (X01 X1 )−1 X01 (X1 θ 1 + X 2 θ2 ) = θ 1 + (X01 X1 )−1 X01 X2 θ 2 = θ 1 + Aθ 2 • When: – θ 1 contains the intercept and main effects – θ 2 contains two-factor interactions – columns of X1 are +1/-1 and orthogonal • A= 1 0X X N 1 2 • Look at this for P-B designs in 8 and 12 runs: 7 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Example, N = 8, elements of A0 I AB AC AD AE AF AG BC BD BE BF BG CD CE CF CG DE DF DG EF EG FG A B C D E F -1 G -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 8 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Look at column for A ... A = -BF = -CD = -EG • ... or, I = -ABF = -ACD = -AEG • But that’s not all ... 9 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Continuing this idea: A = -BF = -CD = -EG I = -ABF = -ACD = -AEG B = -AF = -CG = -DE I = -ABF = -BCG = -BDE C = -AD = -BG = -EF I = -ACD = -BCG = -CEF D = -AC = -BE = -FG I = -ACD = -BDE = -DFG E = -AG = -BD = -CF I = -AEG = -BDE = -CEF F = -AB = -CE = -DG I = -ABF = -CEF = -DFG G = -AE = -BC = -DF I = -AEG = -BCG = -DFG • Generalized interactions of these are: +BCDF, +BEFG, +ACFG, +ADEF, +ABCE, +CDEG, +ABDG, -ABCDEFG • 15 words = 24 − 1 ... this is a 27−4 regular fractional factorial 10 STAT 512 2-Level Factorial Experiments: Irregular Fractions 11 • Example, N = 12, f = 11, elements of 3A0 I A B AB C D E F G H J K L - - - + - - + + - + - - + - + - - + + - - - - - - - - - + + - + - - - + - + - - - + - + AC - AD - + AE - - + AF + - + - AG - + - - - AH - - - - + + AJ + + - - - - - AK + - - + - + - - AL - - - + - - + + - - - - + - - + + + - - + - + - + + - - - - - - - - + BC - BD - - BE - - + BF + - - + ... - STAT 512 2-Level Factorial Experiments: Irregular Fractions • Each main effect is partially aliased with all two-factor interactions of other factors (only) • The weight of each aliased term is 1/3 (rather than 1) • So, for exmple: E(β̂) = β + 13 [−(αγ) − (αδ)...] • There are more two-factor interactions that could bias any main effect, but the bias associated with any one of them is reduced by a factor of 13 . 12 STAT 512 2-Level Factorial Experiments: Irregular Fractions IRREGULAR RESOLUTION IV DESIGNS: • P-B designs are Resolution III because main effects, while orthogonally estimable, would be biased by 2-factor interactions: A= 1 0X X N 1 2 6= 0 • Think about the (complete) foldover of a P-B design (e.g. all runs repeated with all factors reversed). Model matrices for the “doubled” design: X∗1 = 1 D 1 −D X∗2 = X2 X2 13 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Alias matrix: A∗ = 1 2N 1 2N 10 D0 −D0 10 X2 X2 210 X2 D0 X2 − D0 X2 = – 10 X2 = 0 because the P-B design is an OA(2) – D0 X2 − D0 X2 = 0 – therefore A∗ = 0 • So this produces irregular Resolution IV designs in N = the smallest multiple of 8 ≥ 2(f + 1) – e.g. for f = 9 and N = 24 – regular fraction would require 29−4 IV = 32 runs. 14 STAT 512 2-Level Factorial Experiments: Irregular Fractions EXERCISE • Suppose you begin a study with an N -run OA(2) main-effects design (not necessarily a regular fractional factorial). • You decide to augment this initial design with its complete fold-over, e.g. N more runs selected by reversing all factors in all the original runs. • What are the statistical properties of the main effect estimates (sampling variances and possible bias due to two-factor interactions) if the two halves must be treated as blocks? 15 STAT 512 2-Level Factorial Experiments: Irregular Fractions 16 ONE-FACTOR-AT-A-TIME (OAT) DESIGNS • For any f , N = f + 1 • Version 1: − − ... − f + 1 f − 3 f − 3 ... f − 3 + − ... − f − 3 f + 1 f − 3 ... f − 3 D = − + ... − , X01 X1 = f − 3 f − 3 f + 1 ... f − 3 ... ... ... ... ... ... ... ... ... − − ... + f − 3 f − 3 f − 3 ... f + 1 • Version 2: f +1 f −1 f −3 ... −(f − 1) f −1 + − ... − 0 D= + + ... − , X1 X1 = f − 3 ... ... ... ... ... f +1 f −1 f −1 f +1 ... −(f − 3) ... ... − − ... − + + ... + ... −(f − 5) ... −(f − 1) −(f − 3) −(f − 5) ... ... f +1 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Regardless of the value of f , for version 1: – {(X01 X1 )−1 }i,i = – {(X01 X1 )−1 }i,j = 1 2 1 4 (comp. to 1 N for orth. designs) (comp. to 0 for orth. designs) (other than the first row/column) • For version 2: – {(X01 X1 )−1 }i,i = 1 2 – {(X01 X1 )−1 }i,i+1 = − 41 , other {(X01 X1 )−1 }i,j = 0 • Why? In each case β̂i = half the difference of two data values ... • So, much less efficient than orthogonal Resolution III designs • Perhaps reasonable only if σ 2 is known to be very small relative to the size of important effects 17 STAT 512 2-Level Factorial Experiments: Irregular Fractions SUPERSATURATED DESIGNS • N < f + 1 ... full main effects model isn’t estimable • Used in preliminary factor screening, or where operational requirements demand a very small N • Effect sparsity must be “strongly assumed” • A popular example: Lin (Technometrics 1993) – For f factors, begin by constructing a P-B design in f + 1 factors. – Select one column as the “branching column”, include only the runs associated with one level of this factor (i.e. half the number of runs in the P-B plan). – Remove the branching column (which is now confounded with the intercept). 18 STAT 512 2-Level Factorial Experiments: Irregular Fractions • Example: f = 10 • Construct f = 11 P-B design with 12 runs; include only the runs with + in the 11th column: −++−+++−−− − − − + − + + − + + + − − − + − + + − + D= + + − − − + − + + − − + + + − − − + − + +−+++−−−+− • X01 X1 has – diagonal elements = 6 – other row/column 1 elements = 0 – all other elements = ±2 • Can fit models including the mean and any 4 main effects • Stepwise regression is often used for analysis 19 STAT 512 2-Level Factorial Experiments: Irregular Fractions Can you do this with regular fractions of Resolution III (including P-B designs with N a power of 2)? • Yes, but ... • Suppose you start with I = ABC = ... • Use A as a branching column. This is the same as “splitting” again in the regular fractional factorial framework, adding A to the list of factorial effects in the generating relation. • Then I = BC results, so B=C • The column-orthogonality of these designs prevents simultaneous fitting of any small subset of main effects 20 STAT 512 2-Level Factorial Experiments: Irregular Fractions SOME OTHER IRREGULAR FRACTIONAL FACTORIALS: • Full 2f plus a regular fraction – All factorial effects are estimable, some are correlated – Variance of effect estimates is < σ 2 /2f , but > σ 2 /N – Replication without doubling everything • 2f −1 plus an included smaller fraction (e.g. 2f −2 ) – Same sub-comments as above, but for “strings” and “2f −1 ” • 3 distinct 2f −3 ’s from the same generating relation (“3/8 rep”) – More like P-B designs; complex aliasing, more estimable functions 21