The successful use of fractional factorial designs is based on three key ideas: - PowerPoint PPT Presentation

Loading...

PPT – The successful use of fractional factorial designs is based on three key ideas: PowerPoint presentation | free to download - id: 261ccb-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

The successful use of fractional factorial designs is based on three key ideas:

Description:

The successful use of fractional factorial designs is based on three key ideas: The sparsity of effects principle. ... When there are several variables, the ... – PowerPoint PPT presentation

Number of Views:67
Avg rating:3.0/5.0
Slides: 31
Provided by: zaid7
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: The successful use of fractional factorial designs is based on three key ideas:


1
Fractional Factorial
  • The successful use of fractional factorial
    designs is based on three key ideas
  • The sparsity of effects principle. When there are
    several variables, the system or process is
    likely to be driven primarily by some of the main
    effects an low order interactions.
  • The projection property. Fractional factorial
    designs can be projected into stronger designs in
    the subset of significant factors.
  • Sequential experimentation.

2
Fractional Factorial
For a 24 design (factors A, B, C and D) a
one-half fraction, 24-1, can be constructed as
follows
Choose an interaction term to completely
confound, say ABCD.
Using the defining contrast L x1 x2 x3 x4
like we did before we get
3
Fractional Factorial
L x1 x2 x3 x4 mod 2 L x1 x2 x3 x4 mod 2
0000 0 0 0 0 0 0110 0 1 1 0 0
0001 0 0 0 1 1 1010 0 1 0 1 0
0010 0 0 1 0 1 1100 1 1 0 0 0
0100 0 1 0 0 1 0111 0 1 1 1 1
1000 1 0 0 0 1 1011 1 0 1 1 1
0011 0 0 1 1 0 1101 1 1 0 1 1
0101 0 1 0 1 0 1110 1 1 1 0 1
1001 1 0 0 1 0 1111 1 1 1 1 0
4
Fractional Factorial
Hence, our design with ABCD completely confounded
is as follows
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD0 Y00000 Y00011 Y00101 Y01001
ABCD1 Y00001 Y00010 Y00100 Y01000
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD0 Y00110 Y01010 Y01100 Y01111
ABCD1 Y00111 Y01011 Y01101 Y01110
The fractional factorial design
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD1 Y00001 Y00010 Y00100 Y01000
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD1 Y00111 Y01011 Y01101 Y01110
5
Fractional Factorial
Each calculated sum of squares will be associated
with two sources of variation.
Source Prin. Frac. Alias Source Prin. Frac. Alias
A ABCD A2BCD BCD BC ABCD AB2C2D AD
B ABCD AB2CD ACD BD ABCD AB2CD2 AC
C ABCD ABC2D ABD CD ABCD ABC2D2 AB
D ABCD ABCD2 ABC ABC ABCD A2B2C2D D
AB ABCD A2B2CD CD ABD ABCD A2B2CD2 C
AC ABCD A2BC2D BD ACD ABCD A2BC2D2 B
AD ABCD A2BCD2 BC BCD ABCD AB2C2D2 A
6
Fractional Factorial
Lets clean a bit
Source Alias Source Alias
A BCD BC AD
B ACD BD AC
C ABD CD AB
D ABC ABC D
AB CD ABD C
AC BD ACD B
AD BC BCD A
7
Fractional Factorial
Lets reorganize
Complete 23 Design
Source Alias
A BCD
B ACD
C ABD
AB CD
AC BD
BC AD
ABC D
8
Fractional Factorial
So to analyze a 24-1 fractional factorial design
we need to run a complete 23 factorial design
(ignoring one of the factors) and analyze the
data based on that design and re-interpret it in
terms of the 24-1 design.
9
Fractional Factorial
Resolution
  • Many resolutions the three listed in the book
    are
  • Resolution III designs No main effect is aliased
    with any other main effect, they are aliased with
    two factor interactions and two factor
    interactions are aliased with each other. Example
    2III3-1 with ABC as the principle fractions.
  • Resolution IV designs No main effect is aliased
    with any other main effect or any two factor
    interaction, but two factor interactions are
    aliased with each other. Example, 2IV4-1 with
    ABCD as the principle fraction.
  • Resolution V designs. No main effect or
    two-factor interactions is aliased with any other
    main effect or two-factor interaction, but
    two-factor interactions are aliased with three
    factor interactions. Example, 2V5-1 with ABCDE as
    the principle fraction.

10
Fractional Factorial
Example
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD0 3 4 7 2
ABCD0 6 3 6 2
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD0 7 2 5 9
ABCD0 8 3 6 5
11
Fractional Factorial
Assuming all factors are fixed, the linear model
is as follows
12
Fractional Factorial
If we still cant run this design all at once, we
can block that is we can implement a
group-interaction confounding step. We can
confound the highest level interaction of the 23
design, as we did before.
13
Group-Interaction Confounded designs
Partial confounding
Confounding ABC replicate 1
L x1 x2 x3 mod 2
000 0 0 0 0
001 0 0 1 1
010 0 1 0 1
100 1 0 0 1
011 0 1 1 0
101 1 0 1 0
110 1 1 0 0
111 1 1 1 1
14
Group-Interaction Confounded designs
Partial confounding
Confounding ABC replicate 1
a0b0c0 a0b0c1 a0b1c0 a1b0c0 a0b1c1 a1b0c1 a1b1c0 a1b1c1
Block0 Y0000 Y0011 Y0101 Y0110
Block1 Y0001 Y0010 Y0100 Y0111
15
Fractional Factorial
For a 24 design (factors A, B, C and D) a
one-quarter fraction, 24-2, can be constructed as
follows
Choose two interaction terms to confound, say ABD
and ACD, these will serve as our principle
fractions. The third interaction, called the
generalized interaction, that we confounded in
the way is A2BCD2 BC.
Need two defining contrasts L1 x1 x2 0
x4 and L2 x1 0 x3 x4
16
Fractional Factorial
L1 x1 x2 0 x4 mod 2 L2 x1 0 x3 x4 mod 2
0000 0 0 0 0 0 0000 0 0 0 0 0
0001 0 0 0 1 1 0001 0 0 0 1 1
0010 0 0 0 0 0 0010 0 0 1 0 1
0100 0 1 0 0 1 0100 0 0 0 0 0
1000 1 0 0 0 1 1000 1 0 0 0 1
0011 0 0 0 1 1 0011 0 0 1 1 0
0101 0 1 0 1 0 0101 0 0 0 1 1
1001 1 0 0 1 0 1001 1 0 0 1 0
17
Fractional Factorial
L1 x1 x2 0 x4 mod 2 L2 x1 0 x3 x4 mod 2
0110 0 1 0 0 1 0110 0 0 1 0 1
1010 1 0 0 0 1 1010 1 0 1 0 0
1100 1 1 0 0 0 1100 1 0 0 0 1
0111 0 1 0 1 0 0111 0 0 1 1 0
1011 1 0 0 1 0 1011 1 0 1 1 1
1101 1 1 0 1 1 1101 1 0 0 1 0
1110 1 1 0 0 0 1110 1 0 1 0 0
1111 1 1 0 1 1 1111 1 0 1 1 1
18
Fractional Factorial
L1 L2 a b c d L1 L2 a b c d
0 0 0 0 0 0 1 0 0 1 0 0
1 0 0 1 0 0 1 1
0 1 1 1 1 0 1 0
1 1 1 0 1 1 0 1
1 1 0 0 0 1 0 1 0 0 1 0
1 0 0 0 0 1 0 1
0 1 1 0 1 1 0 0
1 1 1 1 1 0 1 1
19
Fractional Factorial
Hence, our design with ABCD completely confounded
is as follows
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD0(00) Y00000 Y01001
ABCD1(11) Y10001 Y11000
ABCD2(01) Y20100 Y20011
ABCD3(10) Y30010 Y30101
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD0(00) Y00111 Y01110
ABCD1(11) Y10110 Y21111
ABCD2(01) Y21010 Y21101
ABCD3(10) Y31100 Y31011
20
Fractional Factorial
One of the possible one-quarter designs is
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD2(01) Y20100 Y20011
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD2(01) Y21010 Y21101
21
Fractional Factorial
Each calculated sum of squares will be associated
with four sources of variation.
Source Prin. Frac. Alias
A ABD,ACD,BC A2BD, A2CD, ABC BD,CD,ABC
B ABD,ACD,BC AB2D, ABCD, B2C AD,ABCD,C
C ABD,ACD,BC ABCD, AC2D, BC2 ABCD,AD,B
D ABD,ACD,BC ABD2, ACD2, BCD AB,AC,BCD
AB ABD,ACD,BC A2B2D, A2BCD, AB2C D,BCD,AC
AC ABD,ACD,BC A2BCD, A2C2D, ABC2 BCD,D,AB
22
Fractional Factorial
Each calculated sum of squares will be associated
with four sources of variation.
Source Prin. Frac. Alias
AD ABD,ACD,BC A2BD2, A2CD2, ABC B,C,ABC
BD ABD,ACD,BC AB2D2, ACD2, B2CD A,AC,CD
CD ABD,ACD,BC ABCD2, AC2D2, BC2D ABC,A,BD
ABC ABD,ACD,BC A2B2CD, A2BC2D, AB2C2 CD,BD,A
BCD ABD,ACD,BC AB2CD2, ABC2D2, B2C2D AC,B,D
ABCD ABD,ACD,BC A2B2CD2, A2BC2D2, AB2C2D C,B,AD
23
Fractional Factorial
The above is not quite satisfactory because we
are aliasing some of the main effects with other
main effects i.e. the resolution is not good
enough!!!
24
Fractional Factorial
What happens after analyzing the data Can do a
confirmatory experiment, complete the block!!
25
Fractional Factorial
L1 x1 x2 0 x4 mod 2 L2 x1 0 x3 x4 mod 2
0000 0000
0001 0001
0010 0010
0100 0100
1000 1000
0011 0011
0101 0101
1001 1001
26
Fractional Factorial
L1 x1 x2 0 x4 mod 2 L2 x1 0 x3 x4 mod 2
0110 0110
1010 1010
1100 1100
0111 0111
1011 1011
1101 1101
1110 1110
1111 1111
27
Fractional Factorial
L1 L2 a b c d L1 L2 a b c d
0 0 1 0



1 1 0 1



28
Fractional Factorial
Hence, our design with ABCD completely confounded
is as follows
a0b0c0d0 a0b0c0d1 a0b0c1d0 a0b1c0d0 a1b0c0d0 a0b0c1d1 a0b1c0d1 a1b0c0d1
ABCD0(00)
ABCD1(11)
ABCD2(01)
ABCD3(10)
a0b1c1d0 a1b0c1d0 a1b1c0d0 a0b1c1d1 a1b0c1d1 a1b1c0d1 a1b1c1d0 a1b1c1d1
ABCD0(00)
ABCD1(11)
ABCD2(01)
ABCD3(10)
29
Fractional Factorial
Each calculated sum of squares will be associated
with four sources of variation.
Source Prin. Frac. Alias
A ABD,ACD,BC
B ABD,ACD,BC
C ABD,ACD,BC
D ABD,ACD,BC
AB ABD,ACD,BC
AC ABD,ACD,BC
30
Fractional Factorial
Each calculated sum of squares will be associated
with four sources of variation.
Source Prin. Frac. Alias
AD ABD,ACD,BC
BD ABD,ACD,BC
CD ABD,ACD,BC
ABC ABD,ACD,BC
BCD ABD,ACD,BC
ABCD ABD,ACD,BC
About PowerShow.com