You are on page 1of 8

STAT 503, Fall 2005 Homework Solution 3 Due: Sept.

22, 2005
4-1 A chemist wishes to test the effect of four chemical aents on the strenth of a particular t!pe of
cloth. "ecause there miht #e $aria#ilit! from one #olt to another, the chemist %eci%es to use a ran%omi&e%
#lock %esin, with the #olts of cloth consi%ere% as #locks. She selects fi$e #olts an% applies all four
chemicals in ran%om or%er to each #olt. The resultin tensile strenths follow. Anal!&e the %ata from this
e'periment (use ) 0.05* an% %raw appropriate conclusions.
"olt
+hemical , 2 3 - 5
, .3 /0 .- ., /.
2 .3 /. .5 .2 .0
3 .5 /0 .0 .3 /0
- .3 ., .5 .5 /1
Design Expert Output
Response: Strength
ANOVA for Selected Factorial Model
Analysis of variance table !artial s"# of s$"ares%
S"# of Mean F
So"rce S$"ares &F S$"are Val"e !rob ' F
"lock ,5..00 - 31.25
2o%el ,2.15 3 -.32 2.30 0.,2,, not sinificant
A 12.95 3 4.32 2.38 0.1211
3esi%ual 2,.00 ,2 ,.02
+or Total ,1,..5 ,1
The 42o%el F5$alue4 of 2.30 implies the mo%el is not sinificant relati$e to the noise. There is a
,2.,, 6 chance that a 42o%el F5$alue4 this lare coul% occur %ue to noise.
Std. Dev. 1.35 R-Squared 0.3727
Mean 71.75 Adj R-Squared0.215
!.". 1. #red R-Squared -0.7$2%
738SS /0.5/ A%e9 7recision ,0.550
(reat#ent Means )Ad*"sted+ ,f Necessary-
.sti#ated Standard
Mean .rror
,5, .0./0 0./0
252 .,.-0 0./0
353 .2.-0 0./0
-5- .2./0 0./0
Mean Standard t for /
0
(reat#ent &ifference &F .rror 1oeff20 !rob ' 3t3
, $s 2 50.00 , 0.05 50.1- 0.3//5
, $s 3 5,.00 , 0.05 52.,, 0.05/-
, $s - 52.00 , 0.05 52.35 0.03.0
2 $s 3 5,.00 , 0.05 5,.,. 0.2/35
2 $s - 5,.20 , 0.05 5,.-, 0.,0-/
3 $s - 50.20 , 0.05 50.23 0.0,05
There is no %ifference amon the chemical t!pes at ) 0.05 le$el.
4-14 Suppose that the o#ser$ation for chemical t!pe 2 an% #olt 3 is missin in 7ro#lem -5,. Anal!&e the
pro#lem #! estimatin the missin $alue. 7erform the e'act anal!sis an% compare the results.
y
23
is missin.
( ) ( )
( ) ( )
( ) ( )
: : :
2. .3 ..
23
- 202 5 22. ,3/0
; .5.25
, , 3 -
ay by y
y
a b
+
+
= = =

35,
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
Therefore, y2.)35..25, y.3)302.25, an% y..),-35.25
For the appro'imate anal!sis, we can replace the missin $alue with the estimate an% fit the mo%el.
Source SS DF MS F0
+hemicals ,2..0-- 3 -.2/,5 2.,5-
"olts ,50.00.5 -
8rror 2,../25 ,, ,.1.0-
Total ,13.-3-- ,0
F0.05,3,,,)3.51, +hemicals are not sinificant. This is the same result as foun% in 7ro#lem -5,.
For the e'act anal!sis, we can fit the <=2 with the %ata (with missin $alue* to et the result.
2inita# output
General Linear Model: Strength versus Bolt, Chemist
Factor Type Levels Values
Bolt fixed 5 1, 2, 3, 4, 5
Chemist fixed 4 1, 2, 3, 4
!alysis of Varia!ce for "tre!#th, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
Bolt 4 14)*215 14+*,3, 35*2+- 1.*,+ +*+++
Chemist 3 12*)54 12*)54 4*21, 2*13 +*154
/rror 11 21*.)2 21*.)2 1*-.,
Total 1, 1,+*)32
" 0 1*4+)5) 12"& 0 ,.*-53 12"&4ad$5 0 ,+*2-3
+hemicals are not sinificant. This is the same result as foun% in the a#o$e appro'imate anal!sis.
Source DF SS(e'act* SS(appro'imate*
+hemicals 3 ,2./5- ,2..0--
"locks - ,-0.030 ,50.00.5
8rror ,, 2,../3 2,../25
Total ,0 ,00./32 ,13.-3--
>ote that for the e'act anal!sis,
E Blocks Chemicalss T
SS SS SS SS + + .
4-4 The effect of three %ifferent lu#ricatin oils on fuel econom! in %iesel truck enines is #ein stu%ie%.
Fuel econom! is measure% usin #rake5specific fuel consumption after the enine has #een runnin for ,5
minutes. Fi$e %ifferent truck enines are a$aila#le for the stu%!, an% the e'perimenters con%uct the
followin ran%omi&e% complete #lock %esin.
Truck
?il , 2 3 - 5
, 0.500 0./3- 0.-0. 0.321 0.5,2
2 0.535 0./.5 0.520 0.-35 0.5-0
3 0.5,3 0.515 0.-00 0.-00 0.5,0
(a* Anal!i&e the %ata from this e'periment.
352
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
From the anal!sis #elow, there is a sinificant %ifference #etween lu#ricatin oils with rear%s to fuel
econom!.
Desin 8'pert ?utput
Response: F"el cons"#ption
ANOVA for Selected Factorial Model
Analysis of variance table (er#s added se$"entially )first to last-%
S"# of Mean F
So"rce S$"ares &F S$"are Val"e !rob ' F
"lock 0.012 - 0.023
2o%el /..0/85003 2 3.35385003 /.35 0.0223 sinificant
A .!0E"003 2 3.353E"003 .35 0.0223
3esi%ual -.22285003 0 5.2.08500-
+or Total 0.,0 ,-
The 2o%el F5$alue of /.35 implies the mo%el is sinificant. There is onl!
a 2.236 chance that a 42o%el F5@alue4 this lare coul% occur %ue to noise.
St%. De$. 0.023 35S9uare% 0./,3/
2ean 0.5, A%A 35S9uare% 0.5,.0
+.@. -.-1 7re% 35S9uare% 50.3503
738SS 0.0,5 A%e9 7recision ,0.0,-
(reat#ent Means )Ad*"sted+ ,f Necessary-
.sti#ated Standard
Mean .rror
,5, 0.-1 0.0,0
252 0.5- 0.0,0
353 0.50 0.0,0
Mean Standard t for /
0
(reat#ent &ifference &F .rror 1oeff20 !rob ' 3t3
, $s 2 50.0-1 , 0.0,5 53.3- 0.0,02
, $s 3 50.00085003 , 0.0,5 50./, 0.5/,5
2 $s 3 0.0-0 , 0.0,5 2..- 0.0255
(#* Bse the Fisher =SD metho% to make comparisons amon the three lu#ricatin oils to %etermine
specificall! which oils %iffer in #reak5specific fuel consumption.
Fisher =SD: =SD) # MSE $
b a
C 2
* , *( , ( , 2 C
when the %esin is #alance%.
For treatment oil of this %ata, we ha$e 2S8)0.00052.0, n,)n2)n3)5, a)3, #)5
So =SD)2.30/D 5 C 00052.0 . 0 D 2 )0.0335
E
. 2 . ,
y y E) 0.5-,050.-12-)0.0-0/F=SD
E
. 3 . ,
y y E) 0.50,2 50.-12-)0.0000G=SD
E
. 3 . 2
y y E) 0.5-,050.50,2)0.0310F=SD
"ase% on the calculation a#o$e, the means for #reak5specific fuel consumption for oils , an% 3 %o not
%ifferH howe$er, oil 2 is %ifferent than oils , an% 3.
(c* Anal!&e the resi%uals from this e'periment.
The resi%ual plots #elow %o not i%entif! an! $iolations to the assumptions.
353
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
Resi dual
N
o
r
m
a
l

%

P
r
o
b
a
b
i
l
i
t
y
Normal Plot of Residuals
-0.0398667 -0.0243167 -0.00876667 0.00678333 0.0223333
1
5
10
20
30
50
70
80
90
95
99
Predited
R
e
s
i
d
u
a
l
s
Residuals !s. Predited
-0.0398667
-0.0243167
-0.00876667
0.00678333
0.0223333
0.37 0.44 0.52 0.59 0.66
"il
R
e
s
i
d
u
a
l
s
Residuals !s. "il
-0.0398667
-0.0243167
-0.00876667
0.00678333
0.0223333
1 2 3
4-14 An in%ustrial enineer is con%uctin an e'periment on e!e focus time. He is intereste% in the effect
of the %istance of the o#Aect from the e!e on the focus time. Four %ifferent %istances are of interest. He has
fi$e su#Aects a$aila#le for the e'periment. "ecause there ma! #e %ifferences amon in%i$i%uals, he %eci%es
to con%uct the e'periment in a ran%omi&e% #lock %esin. The %ata o#taine% follow. Anal!&e the %ata from
this e'periment (use ) 0.05* an% %raw appropriate conclusions.
Su#Aect
Distance (ft* , 2 3 - 5
- ,0 / / / /
/ . / / , /
0 5 3 3 2 5
,0 / - - 2 3
(a* Anal!i&e the %ata from this e'periment.
35-
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
Desin 8'pert ?utput
Response: Foc"s (i#e
ANOVA for Selected Factorial Model
Analysis of variance table !artial s"# of s$"ares%
S"# of Mean F
So"rce S$"ares &F S$"are Val"e !rob ' F
"lock 3/.30 - 1.0.
2o%el 32.15 3 ,0.10 0./, 0.0025 sinificant
A 32.95 3 10.98 8.1 0.0025
3esi%ual ,5.30 ,2 ,.2.
+or Total 0-.55 ,1
The 2o%el F5$alue of 0./, implies the mo%el is sinificant. There is onl!
a 0.256 chance that a 42o%el F5@alue4 this lare coul% occur %ue to noise.
St%. De$. ,.,3 35S9uare% 0./021
2ean -.05 A%A 35S9uare% 0./03/
+.@. 23.20 7re% 35S9uare% 0.,,12
738SS -2.50 A%e9 7recision ,0.-32
(reat#ent Means )Ad*"sted+ ,f Necessary-
.sti#ated Standard
Mean .rror
,5- /.00 0.50
25/ 5.20 0.50
350 3./0 0.50
-5,0 3.00 0.50
Mean Standard t for /
0
(reat#ent &ifference &F .rror 1oeff20 !rob ' 3t3
, $s 2 ,./0 , 0.., 2.2- 0.0--0
, $s 3 3.20 , 0.., -.-0 0.0000
, $s - 3.00 , 0.., -.20 0.00,2
2 $s 3 ,./0 , 0.., 2.2- 0.0--0
2 $s - ,.-0 , 0.., ,.1/ 0.0.3/
3 $s - 50.20 , 0.., 50.20 0..0-2
Distance has a statisticall! sinificant effect on mean focus time.
(#* Fin% the appropriate reression mo%el for the 9uantitati$e factor.
2inita# output
'TB 6 7L' 8Focus Time8 0 "u9$ect %ista!ce:
";BC6 Brief 2 *
General Linear Model: Focus Time versus Subject, Distance
Factor Type Levels Values
"u9$ect fixed 5 1 2 3 4 5
%ista!ce fixed 4 4 ) , 1+
!alysis of Varia!ce for Focus Ti, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
"u9$ect 4 3)*3++ 3)*3++ -*+.5 .*12 +*++4
%ista!ce 3 32*-5+ 32*-5+ 1+*-,3 ,*)1 +*++3
/rror 12 15*3++ 15*3++ 1*2.5
Total 1- ,4*55+
;!usual <9servatio!s for Focus Ti
<9s Focus Ti Fit "/ Fit 1esidual "t 1esid
- 1*++++ 3*1+++ +*.141 22*1+++ 22*4+1
1 de!otes a! o9servatio! =ith a lar#e sta!dardi>ed residual*
'TB 6 7L' 8Focus Time8 0 "u9$ect:
355
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
";BC6 Covariates 8%ista!ce8:
";BC6 Brief 2 *
General Linear Model: Focus Time versus Subject
Factor Type Levels Values
"u9$ect fixed 5 1 2 3 4 5
!alysis of Varia!ce for Focus Ti, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
%ista!ce 1 2,*+-+ 2,*+-+ 2,*+-+ 1-*51 +*++1
"u9$ect 4 3)*3++ 3)*3++ -*+.5 )*3+ +*++4
/rror 14 2+*1)+ 2+*1)+ 1*44+
Total 1- ,4*55+
Term Coef "/ Coef T (
Co!sta!t ,*5)++ +*,,1, -*.1 +*+++
%ista!ce 2+*53++ +*12++ 24*42 +*++1
;!usual <9servatio!s for Focus Ti
<9s Focus Ti Fit "/ Fit 1esidual "t 1esid
- 1*++++ 3*2,++ +*)11- 22*2,++ 22*211
1 de!otes a! o9servatio! =ith a lar#e sta!dardi>ed residual*
'TB 6 7L' 8Focus Time8 0 "u9$ect:
";BC6 Covariates 8%ista!ce8 8%ist?28:
";BC6 Brief 2 *
General Linear Model: Focus Time versus Subject
Factor Type Levels Values
"u9$ect fixed 5 1 2 3 4 5
!alysis of Varia!ce for Focus Ti, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
%ista!ce 1 2,*+-+ .*11, .*11, 5*.4 +*+32
%ist?2 1 4*+5+ 4*+5+ 4*+5+ 3*2. +*+-4
"u9$ect 4 3)*3++ 3)*3++ -*+.5 .*32 +*++3
/rror 13 1)*11+ 1)*11+ 1*23-
Total 1- ,4*55+
Term Coef "/ Coef T (
Co!sta!t 13*51+ 2*,5, 4*.3 +*+++
%ista!ce 22*1+5+ +*,.,3 22*4+ +*+32
%ist?2 +*1125+ +*+)223 1*,1 +*+-4
35/
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
Iith 2inita#:
,. J first fit the 3+"D mo%el (treat #oth #lock an% treatment as fi'e% effect*, the a%Auste% mean
s9uare of error is ,.2.5.
2. Since the su#Aects (#locks* seem to %iffer sinificantl!, in the secon% step, when fittin the first
%eree reression mo%el, J treat the su#Aect as a in%icator $aria#le. "oth %istance an% su#Aect are
sinificant from the test. Since the reression mo%el has less parameters than the 3+"D mo%el,
we ha$e a little #it hiher a%Auste% mean s9uare of error (,.--*.
3. "! plottin the resi%uals from reression $ersus %istance, we can fin% a little cur$ature. To see if
we can ha$e a #etter fit with a 9ua%ratic mo%el, J a%%e% s9uare% %istance in the mo%el. 8$en
thouh the p5$alue for this term is 0.01-, not sinificant at 0.05, the impro$ement in the resi%uals
(A%A 2S ) ,.231* coul% #e sufficicent Austification to also allow the 9ua%ratic mo%el to #e allowe%
as an accepta#le answer.
4-15 An in%ustrial enineer is in$estiatin the effect of four assem#l! metho%s (A% B% C% D* on the
assem#l! time for a color tele$ision component. Four operators are selecte% for the stu%!. Furthermore,
the enineer knows that each assem#l! metho% pro%uces such fatiue that the time re9uire% for the last
assem#l! ma! #e reater than the time re9uire% for the first, rear%less of the metho%. That is, a tren%
%e$elops in the re9uire% assem#l! time. To account for this source of $aria#ilit!, the enineer uses the
=atin s9uare %esin shown #elow. Anal!&e the %ata from this e'periment ( ) 0.05* %raw appropriate
conclusions.
?r%er of ?perator
Assem#l! , 2 3 -
, C),0 D),- A). B)0
2 B). C),0 D),, A)0
3 A)5 B),0 C),, D)1
- D),0 A),0 B),2 C),-
The 2inita# output #elow i%entifies assem#l! metho% as ha$in a sinificant effect on assem#l! time.
2inita# ?utput
General Linear Model
Factor Type Levels Values
<rder ra!dom 4 1 2 3 4
<perator ra!dom 4 1 2 3 4
'ethod fixed 4 B C %
!alysis of Varia!ce for Time, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
35.
STAT 503, Fall 2005 Homework Solution 3 Due: Sept. 22, 2005
'ethod 3 .2*5++ .2*5++ 24*1). 13*,1 +*++4
<rder 3 1,*5++ 1,*5++ )*1). 3*52 +*+,-
<perator 3 51*5++ 51*5++ 1.*1). -*,1 +*+1+
/rror ) 1+*5++ 1+*5++ 1*.5+
Total 15 153*+++
4-64 Suppose that in 7ro#lem -5,1 the enineer suspects that the workplaces use% #! the four operators
ma! represent an a%%itional source of $ariation. A fourth factor, workplace (% % % * ma! #e intro%uce%
an% another e'periment con%ucte%, !iel%in the <raeco5=atin s9uare that follows. Anal!&e the %ata from
this e'periment (use ) 0.05* an% %raw conclusions.
?r%er of ?perator
Assem#l! , 2 3 -
, C),, B),0 D),- A)0
2 B)0 C),2 A),0 D),2
3 A)1 D),, B). C),5
- D)1 A)0 C),0 B)/
2inita# ?utput
General Linear Model
Factor Type Levels Values
'ethod fixed 4 B C %
<rder ra!dom 4 1 2 3 4
<perator ra!dom 4 1 2 3 4
@orAplac ra!dom 4 a 9 c d
!alysis of Varia!ce for Time, usi!# d$usted "" for Tests
"ource %F "e& "" d$ "" d$ '" F (
'ethod 3 -5*5++ -5*5++ 31*,33 3*4. +*1).
<rder 3 +*5++ +*5++ +*1). +*+2 +*--)
<perator 3 1-*+++ 1-*+++ )*333 +*)- +*)1)
@orAplac 3 .*5++ .*5++ 2*5++ +*2. +*,43
/rror 3 2.*5++ 2.*5++ -*1).
Total 15 15+*+++
2etho% an% workplace %o not ha$e a sinificant effect on assem#l! time. Howe$er, there are onl! three
%erees of free%om for error, so the test is not $er! sensiti$e.
350

You might also like