1112sem1 ST5203

NATIONAL UNIVERSITY OF SINGAPORE
EXAlVIINATION
ST5203 EXPERIMENTAL DESIGN
(Semester 1 : AY 2011-2012)
Novernber 2011 - TimeAllowed: 2Hours
INSTRUCTIONSTO CANDIDATES
1. This examination paper contains FIVE (5) questions and comprises FOURTEEN (14)
printedpages (includingthe present page).
2. Candidates must answer ALL questions on this paper. Thetotalmarkfor this paperis 60.
3. Non-programmablecalculatorscan be used.
4. This is anOPEN BOOKexamination.
5. Writedown your matriculationnumber andseat number in thespace below.
Matriculation No: ______________________________________
Seat No: ______________________________
Question Pointsscored Max. points
1 12
2 12
3 12
4 12
5 12
Total
60
1
ST5203
1. (12pts)An agronomist wants toconduct a field trial to compare theyield ofthree varieties
oftomatoin thefollowing 9 plots.
1 2 3
4 5 6
7 8
91
(a) (4pts) Supposethisis theonlyinformationwe have, whichdesigndoyou suggest? How
should the agronomist do therandomization?
(b) (4pts) Suppose we alsoknow thattheplotsinthesamerow will have thesameamount
ofwater irrigation; while the plots in the different rows may have different amount of
water irrigation. The yield oftomato may be affected by variability in the amount of
available moisture. With this information, which design do you suggest? How should
the agronomist do therandomization?
2
ST5203
(c) (4pts) In addition to the informationin (b), we also know that the plots in the same
columnwillhave thesame amountofsunshinej while theplotsin thedifferent columns
may have different amount of sun shine. The yield of tomato may be affected by
variability in the amount ofavailable sun shine. With this extra information, which
design doyou suggest? How should the agronomistdo the randomization?
2. (12pts) The Kenton Company wished to test 4 different package designs for a new break-
fast cereal. Sixteen stores, with approximately equal sales volumes, were selected as the
experimentalunits. Eachstorewas randomlyassignedoneofthepackagedesigns, witheach
package design assigned to 4 stores. Sales, in number ofcases, were observed for the study
period, andthe resultsare recorded in Table 1. Thedatais analyzed using one-way layout.
Package design Sales Samplemean (yd
E'l (y .. _ y.)2
Ij=l 1) 1.
1 11 17 16 14 14.5 21
2 12 10 15 19 14.0 46
3 23 20 18 17 19.5 21
4 27 33 22 26 27.0 62
3
ST5203
(a) (3pts) Fill up the blanks in the ANOVA table below.
Resource Degrees offreedom Sumofsqures Mean ofsquares
Treatment
Error
(b) (3pts) Conductan F test to testthe null hypothesis thatthefour package designs have
thesame sales. Note: F
O
.
05
,3,12 =3.49.
4
ST5203
(c) (3pts) Use the Thkey method to perform multiple comparisons of the four package
designs atthe 0.05 leveL NoteQO.05,4,12=4.20.
(d) (3pts) Supposetheinterest is totest thenullhypothesis thatthemeansalesfor design
1 is the same as the average sales of all four designs. Conduct a t-test to this null
hypothesis ata = 0.05. Note: to.0
25
,12 2.18.
5
ST5203
3. (12pts)Acompanyconducted a 27 full factorial experiment with factors A, B, C,D, E, F, G.
Then,thedataisprocessed byYate'salgorithminRandallthecomputedeffects, including
the estimated grand mean, are stored in the txt file "Effects_output.txt" in standard
order. AssumethatthecomputationsinYate'salgorithmareallcorrect. Read thefollowing
R outputs to answer all thesub-questions.
> # Load data,
> # "Effects_output.txt" contains estimates of effects from Yate's algorithm
> Effects_data = read.table(IEffects_output.txt", header = F)
> Effects = Effects_data$Vl
> # Remove grandmean
> Effects = Effects[2:length(Effects)]
> # Order the effects
> Effects a = abs(Effects)
> Effects_a_order =order(Effects_a)
> # Number of effects
> I =length(Effects)
> # Some possibly relevant information
> cbind(Effects_a_order[(I-8):I], Effects_a[Effects_a_order[(I-8):I]])
[,1] [,2]
[1,] 97 2.043181
[2,] 113 2.119931
[3,] 16 2.434177
[4,] 111 2.456506
[5,] 116 2.800000
[6,] 114 3.000000
[7,] 85 12.000000
[8,] 31 15.000000
[9,] 11 20.000000
6
ST5203
> # Some possibly relevant computational results
> median(Effects_a)
[1] 0.7903197
> median(Effects_a[Effects_a<2.5*1.5*median(Effects_a)])
[1] 0.7883321
(a) (3pts)Using Lenth method andIndividual ErrorRate (IER) criteria, how manyeffects
should bedeclared assignificant ata =0.02? Note: IERo.02,127 =2.38.
(b) (3pts)UsingLenthmethodandExperimental-wiseErrorRate(EER)criteria,howmany
effects should be declared as significant ata 0.02? Note: EER
o
.
02
,127 =4.13.
7
ST5203
(c) (3pts)SixstudentsattendingST5203independentlyconstructedhalfnormalplotsbased
upon the computed effects. Only one student is doing the correct work, which one is
it? Pointout at least one mistakefor each ofthe other plots.
0
N
~
'"
~
~
.,
;::
1!l
::J
~
.J:i
..
'"
0
0
N
'"
III
-
t;
~
.,
1!l
-
0
::J
~
.J:i
..
'"
0
Student 1
0
0
0
#
o 0"
I
-2 -1 0 2
halfnormalquantile
Student3
0
0
0
."cs;tPO 0 0
-
I I I
0.0 O.S 1,0 1.S 2.0 2.S
halfnormalquantile
0
N
Student2
0
~
~
1!l
::J
(5
III
.J:i
IV
'"
-
0
-
'"
0
,..-
0,0 0.5
0
0
~
1,0 1,S 2,0 2.S
halfnormalquantile
0
N
Student4
0
'" t;
~
..
~
"0
III
.J:i
IV
';!
;::
'"
0
-
I I
0.0 0.5
0
0
~
I I
1.0 1.5 2.0 2.5
halfnormalquantile
8
ST5203
Student 5 Student 6
0
N
It)
15
~
~
0
III
S
~
15 "
co
.0
\0
'"
a
0
0
0
,.,
~
0
N
0
~ 0
15
.rg
0
CD
a
~
S
:::J
15
'" .0
'"
\0
a
0 ~
0.0 0.5 1.0 1.5 2.0 2.5 -2 -1 0 2
half normal quantile half normal quantile
(d) (3pts)Withthecorrect half-normal plot in (c), how many effects should bedeclared as
significant? Clearly writedown the name{s) ofyour declared significant effects.
9
ST5203
4. (12pts}Assume that five factors (A, B, C, D and E), each at two levels, are studied with
runs given by thefollowing treatment combinations (I),ade, bd, abe, ede, ae, bee and abed.
(a) (3pts)Writedown the design matrix ofthe experiment.
(b) (3pts}Writedown thedefining contrast subgroup ofthedesign.
10
ST5203
(c) {3pts)Arethereany cleareffectsinthisdesign? If yes, listallcleareffects. If no, explain
why not.
(d) {3pts)Denote the above design as dll consideranother8-rundesign d
2
by switchingthe
signs in all columns ofd
1
Consider d
3
to the augmented design ofd
1
and d
2
, i.e., d
3
has 16 rows. Find a set ofgenerators for d
3
What has been achieved by adding d
2
?
Explain clearly.
11
ST5203
5. (12pts)Replicate2
3
fullfactorialexperimentswithfactors A, BandC 3times. Eachreplica-
tion is arranged in 4 blocks. Theeffects for constructing the blockingschemes for replicates
aredisplayed as follows.
Replicate1: Dl= AC, D2 ABC;
Replicate 2: DI =BC, D2 =ABC;
Replicate 3: Dl AB, D2 =AC.
After collecting the data from the experiment, Yate's algorithm is applied to each ofthe
replicates. Theobtainedeffect estimatesfrom every replicatearedisplayed in thefollowing
tablein standard order.
Rep. 1 Rep. 2 Rep. 3
2.76 3.94 0.59
-0.30 3.36 3.34
8.46 4.96 4.93
-4.09 -1.92 -1.04
2.16 -1.65 -2.29
4.71 4.91 6.17
-0.38 0.64 0.48
-0.26 1.41 2.32
12
ST5203
(a) (3pts)Obtain the estimatesfor all the effects in this experiment.
(b) (3pts)Estimate(j2, the variancefor each individualobservation. Whatarethedegrees
offreedom ofyour estimates.
13
ST5203
(c) (3pts)Construct 95% confidence intervals for effects A and G. Note: to.0
25
,5 = 2.57.
(d) (3pts)Construct 95% confidence intervals for effects AB and BG.
End of Paper
14

1112sem1 ST5203

Uploaded by

Copyright:

Available Formats

1112sem1 ST5203

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1112sem1 ST5203

Uploaded by

Copyright:

Available Formats

NATIONAL UNIVERSITY OF SINGAPORE

You might also like