4IIChapter 4part II Latine Square Andgrecolatine Square Designs2012

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 44

Chapter-4(part-II)

Latine square and Greco-latine square


designs
• It is two-directional blocking design, row-
blocking and column-blocking.
• It handle two known sources of variation
among experimental units and remove them
from experimental error.
• Each treatment is applied randomly only
once in each row and only once in each
column.

1
Model for LS-design

y ijk     i   j   k   ijk
i, j , k  1,2,3,..., p
where y ijk  the observation in row  i and col  k for j th treatment
 i  the i th row effect
 j  the j th treatment effect
 k  the k th column effect
 ijk  error

2
Assumption

 ijk is NID(0,  2 )

• The model is completely additive, there is no


interaction between rows, columns, and
treatments.
• There is only one observation in each cell, so
that only two of the three subscripts i, j, and k
are needed to denote a particular
observation.
3
Hypotheses Testing
• The null and alternative hypotheses in testing
no difference in treatment means are
H 0 : 1   2  ...   t vs
H 1 :  L   M for L  M

• The test statistics is


MStreat
F ~ F( p 1), ( p 1)( p  2)
MSerror

4
• We may also test for no row effect and no
column effect by forming the ratio of MSRows
or MSColumns to MSE. However, because the
rows and columns represent restrictions on
randomization, these tests may not be
appropriate.

5
Analysis of Variance
• The ANOVA consists of partitioning the sum
of squares of total of N  p 2
observations
into components for rows , columns ,
treatments and errors, with their degrees of
freedom
SStotal  SStreat  SSrow  SScol  SSerror
( p  1)  ( p  1)  ( p  1)  ( p  1)  ( p  1)( p  2)
2

6
ANOVA table for LS-design
(single factor experiment)

Source of Sum of Degree of Mean F


Variation Square freedom square
Treatment SS (p-1) MS
Treatment treatment
Row SS (p-1) MS
Row Row MStreat
Column SS (p-1) MS MS Error
Column Col
Error SS (p-1)(p-2) MS
Error Error
Total SS ----
( p 2  1)
Total
7
Computational Formula
p p

y y
2
2
. j. 2 2
y... j 1
i ..
y...
SS treat  i 1
 SS row   2
p p2 p p
p

y
2

2 p p p 2
.. k
y... y...
S total   yijk 
j 1
SS col   2
2

p p i 1 j 1 k 1 p2
SS error  SS total  ( SS row  SS column  SS treat )
where
yi..  i th row total y. j .  j th treatment total
y..k  k th column total y...  grand total

8
Example-1
The data on grain yield of three maize hybrids
(A,B , and D) and a check variety (C)from an
experiment with Latin square design is given
below. Test the effect of variety of maize on
yield at α=0.05

9
Data on yield
Column
ROW

1 2 3 4

1 1.64(B) 1.21(D) 1.425(C) 1.345(A)

2 1.475(C) 1.185(A) 1.4(D) 1.29(B)

3 1.67(A) 0.710(C) 1.665(B) 1.18(D)

4 1.565(D) 1.290(B) 1.655(A) 0.66(C)

10
solution

1. H 0 :  A   B   C   D
H 1 :  k   m for k  m
2.   0.05
3.Calulate the test statistic
SS treat
MS treatment ( p  1)
F 
MS error SS error
( p  1)( p  2)

11
Treat. A B C D Grand
total
total 5.855 5.885 4.27 5.355 21.365

Row 1 2 3 4 Grand
total
total 5.62 5.35 5.225 5.17 21.365

Column 1 2 3 4 Grand
total
total 6.35 4.395 6.145 4.475 21.365

12

2
5.855 2  5.885 2  4.27 2  5.355 2 21.365 2
2
y y...
     0.42684
. j.
SS treat 2
p p 4 16
t

y
2

5.62 2  5.35 2  5.225 2  5.17 2


2
j 1
i ..
y... 21.365 2
SS row   2    0.03015
p p 4 16
t

 y .. k
2

6.35  4.395 2  6.145 2  4.475 2 21.365 2


2 2
j 1 y...
SS col   2    0.82734
p p 4 16

13
t t t 2
y...
  yijk  2
2
S total
i 1 j 1 k 1 p
21.3652
SS total  1.64  1.475  ...  0.660 
2 2 2

16
SS total  1.41392
SS error  SStotal  ( SS row  SS column  SS treat )
sserror  0.12958

14
ANOVA-table

Source DF SS MS F
Col 3 0.82734 0.27578 12.77
Row 3 0.03015 0.01005
Treat 3 0.42684 0.14228
Error 6 0.12958 0.02160
Total 15 1.41392

15
4. Calculated F=6.59 is greater than tabulated F
, the null hypothesis is rejected .

F0.05,3,6  4.76
5. The varieties have significant effect on
yield of maize at 0.05 level of significance.
(treatment means differ significantly at 0.05
level of significance.)
16
The significance of blocking in RCBD and the
significance of row-blocking and column-
blocking in latine square design can be tested
by comparing F value for blocking, F value for
row, F value for column with corresponding
tabulated F values.
Examples: From the above ANOVA table

17
MS row MS col
Frow   0.47 , Fcol   12.77,
MS error MS error
• The difference among column blocking is
significant, column blocking is successful in
reducing experimental error.
Example -2
An industrial engineer is investigated the effect
of four assembly methods (A,B,C,D)on the
assembly time for a color TV component.

18
Four operators are selected for the study
.Further more, trend develops in the required
assembly time. To account order of assembly
as source of variability ,the engineer uses the
latine square design shown below.
Analyze the data at α=0.05 level of significance.

19
Order of Operators
assembly
1 2 3 4
1 C(10) D(14) A(7) B(8)
2 B(7) C(18) D(11) A(8)
3 A(5) B(10) C(11) D(9)
4 D(10) A(10) B(12) C(14)

20
Solution
1. H 0 :  A   B   C   D
H1 :  k   m for k  m
2.   0.05
3.Calulate the test statistic
SS treat
MS treatment ( p  1)
F  
MS error SS error
( p  1)( p  2)

 y. j .
2 2
y...
SS treat  
p p2
SS error  SS total  ( SS row  SS column  SS treat )

21
treatment A B C D Grand
total
Treatment 30 37 53 44 164
total

Row 1 2 3 4 grand
total

Row total 39 44 35 46 164

Column 1 2 3 4 Grand
total
Column 32 52 41 39 164
total

22
t

y
2

39 2  44 2  35 2  46 2
2
j 1
i ..
y... 164 2
SS row   2    18.5
p p 4 16

y
2

32  52 2  412  39 2 164 2
2 2
.. k
j 1 y
SS col   ...2    51.5
p p 4 16

 y. j .
2
30 2  37 2  53 2  44 2 164 2
2
y...
SS treat   2
   72.5
p p 4 16

23
p p p 2
y...
S total   yijk 
2
2
i 1 j 1 k 1 p
2
164
SS total  10  7  ...  14 
2 2 2
 153
16
SS error  SS total  ( SS row  SS column  SS treat )  10.5

24
ANOVA -table
Source of Sum of Degree of Mean F
variation squares freedom squares
treatment 72.5 3 24.17

row 18.5 3

column 51.5 3 13.81

Error 10.5 6 1.75

total 153 15

25
4.Since calculated value of F is greater than
tabulated value F0.05,3,6  4.76
the null hypothesis is rejected.
5.The four assembly methods have effect on
assembly time at α=0.05 level of significance.

26
The Graeco-Latin Square Design

• The Graeco-Latin square design can be used to


control systematically three sources of
extraneous variability and to block in three
directions.
• The design allows investigation of four factors
(rows, columns, Latin letters, and Greek
letters), each at p levels in only p 2 runs.

27
Model for the Graeco-Latin square
design
• y ijk     i   j   k   l   ijkl
i, j , k , l  1,2,3,..., p
where y ijkl  the observation in row  i , col  k and greeklette r  l
for j th treatment
 i  the i th row effect
 j  the j th treatment effect
 k  the k th column effect
 l  the l th greek letter effect
 ijkl  error term

28
Assumption

 ijkl is NID(0,  ) 2

• The model is completely additive, there is no


interaction between rows, columns, Greek
letters and treatments.
• There is only one observation in each cell, so
that only two of the four subscripts i, j, k and l
are needed to denote a particular
observation.
29
 The Greek letters appear exactly once in each
row and column and exactly once with each
treatment.
 The factor represented by the Greek letters is
orthogonal to rows, columns, and treatments.

30
• The null and alternative hypotheses in testing
no difference in treatment means are
H 0 : 1   2  ...   p vs
H1 :  L   M for L  M

• The test statistic is


MStreat
F ~ F( p 1), ( p 1)( p 3)
MSerror

31
Analysis of variance
• The ANOVA consists of partitioning the sum
of squares of total of N  p 2 observations
into components for rows , columns ,
treatments, greek letters and errors.
SStotal  SStreat  SSrow  SScol  SSGree  SSerror
.

32
ANOVA table for Greco-LatineSquareD.
Source of Sum of Degree of Mean F
variation squares freedom squares

treatment SS (p-1) MS
treatment treatment

row SS (p-1) MS
row row
column SS (P-1) MS MStreat
column column MS error
greekL SS (p-1) MS
greekL greekl
error SS (p-1)(P-3) MS
error error
total SS N-1 ---
total
33
Computational Formula
p

 y. j..
2
2
y....
SS treat  i 1

p p2
p

y
2

2
j 1
i .. . y....
SS row  
p p2
p

y
2

. 2
j 1
.. k
y....
SS col  
p p2

34
p

 y...l
2
2
y....
SSgreeklette  i 1

p p2
p p p p 2
y....
SS total   y ijkl 
2

i 1 j 1 k 1 l 1 p2
SS error  SS total  ( SS row  SS column  SS treat  SS greekl )
y i...  i th row total y. j ..  j th treatment total
y..k .  k th column total y...l  l th greek letter total
y....  grand total

35
Example-1
An industrial engineer is investigated the effect
of four assembly methods (A,B,C,D)on the
assembly time for a color TV component.
Four operators are selected for the study
Further more, trend develops in the required
assembly time. To account order of assembly
and workplace as source of variability ,the
engineer uses the Greco-latine square design
shown below. Analyze the data at α=0.05

36
Order of Operators
assembly
1 2 3 4

1 Cη=11 Bλ=10 Dν=14 Aθ=8

2 Bθ=8 Cν=12 Aλ=10 Dη=12

3 Aν=9 Dθ=11 Bη=7 Cλ=15

4 Dλ=9 Aη=8 Cθ=18 Bν=6

37
Solution
1. H 0 :  A   B   C   D
H 1 :  k   m for k  m
2.   0.05
3.Calulate the test statistic
SS treat
MS treatment ( p  1)
F 
MS error SS error
( p  1)( p  3)
SS error  SS total  ( SS row  SS column  SS treat  SS greekL )

38
Treat. A B C D total

total 35 31 56 46 168

Greekl θ λ η ν total

total 45 44 38 41 168

39
row 1 2 3 4 total

total 43 42 42 41 168

column 1 2 3 4 total

total 37 41 49 41 168

40
 y. j..
2
35 2  312  56 2  46 2 168 2
2
y....
SS treat   2
   95.5
p p 4 16
t

y
2

43 2  42 2  42 2  412 168 2
2
j 1
i ...
y....
SS row   2    0.5
p p 4 16
t

y
2

37  412  49 2  412 168 2


2 2
.. k .
j 1 y....
SS col   2    19
p p 4 16

41
p

 y...l
2
2
y.... 45 2  44 2  38 2  412 168 2
SSgreekL  i 1
    7.5
p p2 4 16
p p p p 2
y...
SS total   y ijkl 
2

i 1 j 1 k 1 l 1 p2
2
168
SS total  112  8 2  ...  6 2   150
16
SS error  SS total  ( SS row  SS column  SS treat  SS greekL )
sserror  27.5

42
ANOVA-table
Source of Sum of Degree of Mean F
variation squares freedom squares

Treatment 95.5 3 31.83

row 0.5 3

column 19 3 3.47

greekl 7.5 3

error 27.5 3 9.167

total 150 15

43
4.Since calculated value of F is less than
tabulated value the null hypothesis is not
rejected.(F0.05,3,3  9.28)
5.The four assembly methods have no significant
effect on assembly time at α=0.05.

44

You might also like