Professional Documents
Culture Documents
ETM501& Applied&Statistics&And& Design&&&Analysis&Of&Experiments&&&
ETM501& Applied&Statistics&And& Design&&&Analysis&Of&Experiments&&&
ETM501&
APPLIED&STATISTICS&AND&
DESIGN&&&ANALYSIS&OF&EXPERIMENTS&&&
&
Ali&R.&Kaylan&
kaylan@boun.edu.tr&
&
Boğaziçi&University&
Department&of&Industrial&Engineering&
Bebek,&İstanbul,&Turkey&
&
Lec$4.$Comparison$of$K$Populations$and$
Kaylan$/$ETM501$ 1$
Block$Designs$
COMPARISON&OF&K[POPULATIONS&AND&BLOCK&DESIGNS&
OUTLINE((
&
I.&Completely&Randomized&Designs&
&
II.&Model&Adequacy&Checking&
&
III.&Tensile&Strength&Experiment&
&
IV.&Randomized&Block&Design&
&
V.&Hardness&TesZng&Experiment&
&
VI.$Latin$Square$Design&
Lec$4.$Comparison$of$K$Populations$and$
Kaylan$/$ETM501$ 2$
Block$Designs$
What&if&there&are&more&than&two&factor&levels?&
The&t[test&does¬&directly&apply.&
&
There&are&lots&of&pracZcal&situaZons&where&there&are&
either&more&than&two&levels&of&interest,&or&there&are&
several&factors&of&simultaneous&interest.&
&
The&analysis&of&variance&(ANOVA)&is&the&appropriate&
analysis& engine &for&these&types&of&experiments.&
The&ANOVA&was&developed&by&Fisher&in&the&early&1920s,&
and&iniZally&applied&to&agricultural&experiments.&
&
Used&extensively&today&for&industrial&experiments.&
&
The&name& analysis&of&variance &stems&from&a&
parZZoning&of&the&total&variability&in&the&response&
variable&into&components&that&are&consistent&with&a&
model&for&the&experiment&
(
Lec 4. Comparison of K Populations
Kaylan / ETM501
and Block Designs 4
EXPERIMENTS&TO&COMPARE&k&AVERAGES&
QuesZon:&Do&all&the&techniques&give&essenZally&the&same&results?&
# H 0 : µA = µB = µC = µD
"
!H1 : At least one mean differs
EXAMPLE(1((Montgomery):(
Three&brands&of&baeeries&are&under&study.&It&is&suspected&that&the&
lives&(in&weeks)&of&the&three&brands&are&different.&Five&baeeries&of&
each&brand&are&tested&with&the&following&results.&
& Brand 1 Brand 2 Brand 3
108 96 74
100 92 83
98 91 79
94 96 79
100 100 75
TOTAL 500 475 390
MEAN 100 95 78
S.Square 104 52 52
H 0 : σ 12 = σ 22
H 1 : σ 12 ≠ σ 22
104
F= = 2 < F0.05, 4, 4 = 6.39
52
22 ± 3.06 * 2.63
22 ± 8.05
13.95 ≤ µ1 − µ 3 ≤ 30.05
EXAMPLE(1(d)(Which&brand&would&you&select&for&use?&If&the&
manufacturer& will& replace& without& charge& any& baeery& that&
fails& in& less& than& 85& weeks,& what& percentage& would& he&
expect&to&replace?&
Select&Brand&1&(longest&life).&&
Assuming&normality,&
( 85 − 100 %
Φ&& ## = Φ(−3.603) = 0.0002
' 17.333 $
EXAMPLE(2.(Comparison&of&four&Techniques&(A,&B,&C,&D)&
Design&Issues:&24&Experiments&with&4,&6,&6,&8&replicaZons&
respecZvely.&In&what&order&to&carry&out&experiments?&
TECHNIQUE
A B C D
20 12 16 23
2 9 7 3
11 15 1 6
10 14 17 18
4 13 22
8 21 19
5
24
Numbers&in&the&layout&designate&the&order&of&the&tests.&&
Lec 4. Comparison of K Populations
Kaylan / ETM501
and Block Designs 11
COMPLETELY&RANDOMIZED&DESIGN&
The(underlying(mathemaCcal(model:(((
( Y = µ + τ + ε
ij j ij
Y = µ + ε
ij j ij
where&
&Yij&=&ith&observaZon&in&group&j&
&µj&=&True&mean&of&group&j&
&εij&=&Error&
&&
A B C D
62 63 68 56
60 67 66 62
63 71 71 60
59 64 67 61
65 68 63
66 68 64
63
59
GROUP AVER. 61 66 68 61
GROUP VAR. 3,33 8,00 2,80 6,86
N 4 6 6 8
WITHIN S.S. 10 40 14 48
GRAND AVER. 64
GROUP DEV. -3 2 4 -3
CalculaCon(of(Variance(:(
Within(groups:(
( k nj
2
( y
∑ ∑ ij - y j )
2 j =1 i =1
sw =
N −k
Between(groups:( nj
k
2
(y
∑∑ j - y )
2 j =1 i =1
s =
b
k −1
While&sums&of&squares&cannot&be&directly&compared&to&test&the&
hypothesis&of&equal&means,&mean&squares&can&be&compared.&
&
A&mean&square&is&a&sum&of&squares÷d&by&its°rees&of&
freedom.&
&
If&Ho&is&true&(the&treatment&means&are&equal),&the&treatment&and&
error&mean&squares&will&be&(theoreZcally)&equal.&&
&
If&H1&is&true&(treatment&means&differ),&the&treatment&mean&square&
will&be&larger&than&the&error&mean&square.&
(
F3,20,0.01&=&4.94,&F3,20,0.05&=&3.10&
&
Do¬&accept&Ho&at&5%&level&of&significance.&
&
62 63 68 56 1 -3 0 -5
60 67 66 62 -1 1 -2 1
63 71 71 60 2 5 3 -1
59 64 67 61 = 64 + -3 2 4 -3 + -2 -2 -1 0
65 68 63 -1 0 2
66 68 64 0 0 3
63 2
59 -2
The(AddiCve(property(of(sum(of(squares:(((
(
D=T-M
k nj k nj
2 2 2
(y
∑∑ ij - y) = y
∑∑ ij - N y
j=1 i =1 j=1 i =1
D = G+ R
k nj k k nj
2 2 2
∑ ∑(y ij - y) = ∑n j (y j - y ) + ∑ ∑ (y ij - yj)
j =1 i =1 j= 1 j= 1 i= 1
SST = SSTreatments + SS E
k nj k k nj
2 2 2
(y
∑∑ ij - y) = n (y
∑ j j - y) + (y -
∑∑ ij jy )
j=1 i =1 j=1 j=1 i =1
SSTreatments&&&:&&A&large&value&reflects&large&differences&in&treatment&
means.&
A&small&value&likely&indicates&no&differences&in&treatment&means.&
ObservaCon(=(Mean(+(Treatment(Effect(+(Residual(
(((((((((((T((((((((=((((((M(((+(((D(
(((((((((((((T((((((((=((((((M(((+(((((((((((((G(((((((((((((((+(((((((R(
Orthogonality(:(((
(
D⊥M G ⊥R M ⊥R M ⊥G
(Checking(AssumpCons:(
(
& & &Normality&
& & &Constant&variance&
& & &Independence&
&
Have%we%fit%the%right%model?%
%
(
Lec 4. Comparison of K Populations
Kaylan / ETM501
and Block Designs 21
MODEL&ADEQUACY&
Residual(Analysis:(
(InvesCgate(if(there(are(any(apparent(abnormaliCes.(
(
1. Residual(Dot(diagrams(
(
2.(((Residuals(by(techniques(
(
3.(((Residuals(in(Cme(sequence(
(
(
EXAMPLE(3.((Montgomery,(3.10(p.132):(
A&product&development&engineer&is&interested&in&invesZgaZng&
the&tensile&strength&of&a&new&syntheZc&fiber&that&will&be&used&to&
make&cloth&for&men’s&shirts.&The&engineer&knows&from&previous&
experience&that&the&strength&is&affected&by&the&weight&percent&
of&coeon&used&in&the&blend&of&materials&for&the&fiber.&
Furthermore,&he&suspects&that&increasing&the&coeon&content&will&
increase&the&strength,&at&least&iniZally.&He&also&knows&that&
coeon&content&should&range&between&about&10&percent&and&40&
percent&if&the&final&cloth&is&to&have&other&quality&characterisZcs.&
The&engineer&decides&to&test&specimens&at&five&levels&of&coeon&
weight&percent:&15&percent,&20&percent,&25&percent,&30&percent,&
and&35&percent.&He&also&decides&to&test&five&specimens&at&each&
level&of&coeon&content.&
a=&5&levels&
n=&5&replicates&&
25&runs&should&be&made&in&random&order.&
2
Observed Tensile Strength (lb/in )
% Weight 1 2 3 4 5 Totals (yi ) Aver.
15 7 7 15 11 9 49 9,80
20 12 17 12 18 18 77 15,40
25 14 18 18 19 19 88 17,60
30 19 25 22 19 23 108 21,60
35 7 10 11 15 11 54 10,80
Does&changing&the&coeon&
weight&percent&change&the&
mean&tensile&strength?&
Is&there&an&opCmum&level&for&
coeon&content?&
Reference&DistribuZon:&
&Response:&Strength&
&((((((((ANOVA(for(Selected(Factorial(Model&
&Analysis(of(variance(table([ParCal(sum(of(squares]&
& &Sum(of & &Mean &F &&
&Source&&&&Squares &DF &Square &Value &Prob(>(F&
&Model &475.76 &4 &118.94 &14.76 &<&0.0001&
&A &475.76 &4 &118.94 &14.76 &<&0.0001&
&Error &161.20 &20 &8.06&
&CorTotal &636.96 &24&
&
&
&Std.&Dev. &2.84 & &R[Squared &0.7469&
&Mean &15.04 & &Adj&R[Squared &0.6963&
&C.V. &18.88 & &Pred&R[Squared &0.6046&
&PRESS &251.88 & &Adeq&Precision &9.294&
ExaminaCon(of(residuals( 99
Design^Expert(generates( 95
the(residuals.(
Norm al % probability
90
80
( 70
Residual(plots(are(very( 50
useful.( 30
20
( 10
Normal(probability(plot(of(
1
residuals.(
Res idual
2.95
2.95
2
Res iduals
Res iduals
2
0.7
0.7
2
2
-1.55
-1.55
2
2
2 -3.8
-3.8
1 4 7 10 13 16 19 22 25
9.80 12.75 15.70 18.65 21.60
The(ANOVA(tests(the(hypothesis(of(equal(treatment(means.(If(
that(hypothesis(is(rejected,(we(don t(know(which(specific(means(
are(different.(
(
(Assume(that(the(model(is(valid.((Residual(analysis(yields(
saCsfactory(results.)(
(
Determining(which(specific(means(differ(following(an(ANOVA(is(
called(the(mulCple(comparisons(problem.(
(
(Pairwise(t^tests(on(means((Also(called(Fisher s(Least(Significant(
Difference(Method(or(Fisher s(LSD).(
&(Treatment(Means((Adjusted,(If(Necessary)&
& &EsCmated & &Standard&
& &Mean & &Error&
&&1[15 &9.80 & &1.27&
&&2[20 &15.40 & &1.27&
&&3[25 &17.60 & &1.27&
&&4[30 &21.60 & &1.27&
&&5[35 &10.80 & &1.27&
&
& &Mean & &Standard &t(for(H0 &&
&Treatment &Difference &DF &Error &Coeff=0 &Prob(>(|t|&
&&&1&vs&&2 &[5.60 &1 &1.80 &[3.12 &0.0054&
&&&1&vs&&3 &[7.80 &1 &1.80 &[4.34 &0.0003&
&&&1&vs&&4 &[11.80 &1 &1.80 &[6.57 &<&0.0001&
&&&1&vs&&5 &[1.00 &1 &1.80 &[0.56 &0.5838&
&&&2&vs&&3 &[2.20 &1 &1.80 &[1.23 &0.2347&
&&&2&vs&&4 &[6.20 &1 &1.80 &[3.45 &0.0025&
&&&2&vs&&5 &4.60 &1 &1.80 &2.56 &0.0186&
&&&3&vs&&4 &[4.00 &1 &1.80 &[2.23 &0.0375&
&&&3&vs&&5 &6.80 &1 &1.80 &3.79 &0.0012&
&&&4&vs&&5 &10.80 &1 &1.80 &6.01 &<&0.0001&
&
&
Tensile&strength&averages&from&the&coeon&weight&percentage&experiment&in&
relaZon&to&a&t&distribuZon&with&a&scale&factor&&
& &(MSE/n)0.5=(8.06/5)0.5=&1.27&
Blocking(Principle:(
(Blocking&is&a&technique&for&dealing&with&nuisance&factors.(
(
• A&nuisance(factor&is&a&factor&that&probably&has&some&affect&on&
the&response,&but&it s&of&no&interest&to&the&experimenter…&
However,&the&variability&it&transmits&to&the&response&needs&to&
be&minimized.&
• Typical&nuisance&factors&include&batches&of&raw&material,&
operators,&pieces&of&test&equipment,&Zme&(shiws,&days,&etc.),&
different&experimental&units.&
• Many&industrial&experiments&involve&blocking.&
• Failure&to&block&is&a&common&flaw&in&designing&an&experiment&
(Consequences?)&
Lec 4. Comparison of K Populations
Kaylan / ETM501
and Block Designs 36
RANDOMIZED&BLOCK&DESIGN&
Blocking(Principle:(
(If&the&nuisance&variable&is&&
& &known&and&controllable,&we&use&blocking. ( ((
( (known(and(uncontrollable,&&
& & &someZmes&we&can&use&the&analysis(of(covariance&to&
remove&the&effect&of&the&nuisance&factor&from&the&analysis.&
( (unknown(and(uncontrollable&(a& lurking (variable),&&
& & &we&hope&that&randomizaCon&balances&out&its&impact&
across&the&&experiment&
&
SomeZmes&several&sources&of&variability&are&combined&in&a&block,&
so&the&block&becomes&an&aggregate&variable&
A B C D E SUM
I 16,9 18,2 17 15,1 18,3 85,5
II 16,5 19,2 18,1 16.,0 18,3 88,1
III 17,5 17,1 17,3 17,8 19,8 89,5
SUM 50,9 54.5 52,4 48,9 56,4 263,1
The(underlying(mathemaCcal(model:(((
( Y = µ + βi + τ + ε
ij j ij
where&
&Yij&=&ith&observaZon&in&group&j&
&µ=&Grand&Mean&
&βi=&Block&i&&
&τj=&Treatment&j& &&
&εij&=&Error&
&&
2
& ε ij : IIDN (0, σ )
DecomposiZon&of&the&Sums&of&squares&:&&
&
SST = SSTreatments + SS Blocks + SS E
Degrees&of&freedom&decomposiZon&are&as&follows:&
&
&
ab − 1 = a − 1 + b − 1 + (a − 1)(b − 1)
RaZos&of&sums&of&squares&to&their°rees&&of&freedom&result&
in&mean&squares.&
RaZo&of&the&mean&square&for&treatments&to&the&error&mean&
square&is&an&F&staZsZc&that&can&be&used&to&test&the&
hypothesis&of&equal&treatment&means.&
ANOVA(TABLE(:(((
F4,8,.05&=&3.84,&F4,8,.10&=&2.81&
&
Treatment&effect&is&significant&at&10%&but¬&at&5%&level.&
1 -2 -1 1 5 3
2 -1 -2 3 4 4
3 -3 -1 0 2 -2
4 2 1 5 7 15
y.j -4 -3 9 18 20
Design-Expert Output
(Response:&&Hardness&
&((((((((ANOVA(for(Selected(Factorial(Model&
&Analysis(of(variance(table([ParCal(sum(of(squares]&
& &Sum(of & &Mean &F &&
&Source &Squares &DF &Square &Value &Prob(>(F&
&Block &82.50 &3 &27.50&
&Model &38.50 &3 &12.83 &14.44 &0.0009&
&A &38.50 &3 &12.83 &14.44 &0.0009&
&Residual &8.00 &9 &0.89&
&Cor&Total&129.00 &15&
&
&Std.&Dev. &0.94 & &R[Squared &0.8280&
&Mean &1.25 & &Adj&R[Squared &0.7706&
&C.V. &75.42 & &Pred&R[Squared &0.4563&
&PRESS &25.28 & &Adeq&Precision &15.635&
Lec 4. Comparison of K Populations
Kaylan / ETM501
and Block Designs 49
HARDNESS&TESTING&EXPERIMENT&
Design-Expert
DE S IG N-E X P E RT P l o t Output – Residual Analysis
Normal plot of residuals
Ha rd n e ss
99
95
Norm al % probability
90
80
70
50
30
20
10
Res idual
0.875 0.875
Res iduals
Res iduals
0.25 2 0.25
-0.375 -0.375
-1
-1
Used&to&simultaneously&control&(or&eliminate)&two(
sources(of(nuisance(variability.&
• &A&significant&assumpZon&is&that&the&three&factors&
(treatments,&nuisance&factors)&do(not(interact.&
&
• &If&this&assumpZon&is&violated,&the&LaZn&square&design&
will¬&produce&valid&results.&
&
• &LaZn&squares&are¬&used&as&much&as&the&RCBD&in&
industrial&experimentaZon.&&
Position (j)
Run(i) 4 2 1 3
II A B D C
III D C A B
I C D B A
IV B A C D
Position (j)
Run(i) 4 2 1 3 Sum Aver.
II 251 241 227 229 948 237
III 234 273 274 226 1007 251,75
I 235 236 218 268 957 239,25
IV 195 270 230 225 920 230
Sum 915 1020 949 948 3832
Aver. 228,75 255 237,25 237 239,5
MATHEMATICAL MODEL :
Y = µ + βi + τ + ηk + ε
ijk j ijk
ANOVA TABLE :
F3,6,.05&=&4.76&
&