1

Chapter 4 Randomized Blocks, Latin

Squares, and Related Designs
2
4.1 The Randomized Complete
Block Design
Nuisance factor: a design factor that probably has
an effect on the response, but we are not interested
in that factor.
If the nuisance variable is known and
controllable, we use blocking
If the nuisance factor is known and
uncontrollable, sometimes we can use the
analysis of covariance (see Chapter 14) to
remove the effect of the nuisance factor from the
analysis
3
If the nuisance factor is unknown and
uncontrollable (a lurking variable), we hope
that randomization balances out its impact across
the experiment
Sometimes several sources of variability are
combined in a block, so the block becomes an
aggregate variable
4
We wish to determine whether 4 different tips
produce different (mean) hardness reading on a
Rockwell hardness tester
Assignment of the tips to an experimental unit;
that is, a test coupon
Structure of a completely randomized experiment
The test coupons are a source of nuisance
variability
Alternatively, the experimenter may want to test
the tips across coupons of various hardness levels
The need for blocking

5
To conduct this experiment as a RCBD, assign all
4 tips to each coupon
Each coupon is called a block; that is, its a
more homogenous experimental unit on which to
test the tips
Variability between blocks can be large,
variability within a block should be relatively
small
In general, a block is a specific level of the
nuisance factor
A complete replicate of the basic experiment is
conducted in each block
A block represents a restriction on
randomization
All runs within a block are randomized
6
Suppose that we use b = 4 blocks:







Once again, we are interested in testing the
equality of treatment means, but now we have to
remove the variability associated with the
nuisance factor (the blocks)
7
Statistical Analysis of the RCBD
Suppose that there are a treatments (factor levels)
and b blocks
A statistical model (effects model) for the RCBD
is


is an overall mean, t
i
is the effect of the ith
treatment, and |
j
is the effect of the jth block
c
ij
~ NID(0,o
2
)

1, 2,...,
1, 2,...,
ij i j ij
i a
y
j b
t | c
=

= + + +

=

0 , 0
1 1
= =

= =
k
j
j
a
i
i
| t
8
Means model for the RCBD

The relevant (fixed effects) hypotheses are

An equivalent way for the above hypothesis

Notations:

j i ij ij ij ij
y | t c + + = + = ,
0 1 2
1
: where (1/ ) ( )
b
a i i j i
j
H b t | t
=
= = = = + + = +

0 :
2 1 0
= = = =
a
H t t t
N y y a y y b y y
y y y y
b j y y
a i y y
j j i i
a
i
i
b
j
j
a
i
b
j
ij
a
i
ij j
b
j
ij i
/ , / , /
,..., 1 ,
,..., 1 ,
1 1 1 1
1
1

=

=

= =

=

= = =
= = =
= =
= =

9
ANOVA partitioning of total variability:
2
.. . .. . ..
1 1 1 1
2
. . ..
2 2
. .. . ..
1 1
2
. . ..
1 1
( ) [( ) ( )
( )]
( ) ( )
( )
a b a b
ij i j
i j i j
ij i j
a b
i j
i j
a b
ij i j
i j
T Treatments Blocks E
y y y y y y
y y y y
b y y a y y
y y y y
SS SS SS SS
= = = =
= =
= =
= +
+ +
= +
+ +
= + +

10
SS
T
= SS
Treatment
+ SS
Blocks
+ SS
E

Total N = ab observations, SS
T
has N 1 degrees
of freedom.
a treatments and b blocks, SS
Treatment
and SS
Blocks

have a 1 and b 1 degrees of freedom.
SS
E
has ab 1 (a 1) (b 1) = (a 1)(b 1)
degrees of freedom.
From Theorem 3.1, SS
Treatment
/o
2
, SS
Blocks
/ o
2
and
SS
E
/ o
2
are independently chi-square
distributions.
11
The expected values of mean squares:







For testing the equality of treatment means,
2
1
2
2
1
2
2
) (
1
) (
1
) (
o
|
o
t
o
=

+ =

+ =

=
=
E
b
j
j
Blocks
a
i
i
Treatment
MS E
b
a
MS E
a
b
MS E
) 1 )( 1 ( , 1 0
~

=
b a a
E
Treatments
F
MS
MS
F
12
The ANOVA table






Another computing formulas:
Blocks Treatments T E
b
j
j Blocks
a
i
i Treatments
a
i
b
j
ij T
SS SS SS SS
N
y
y
a
SS
N
y
y
b
SS
N
y
y SS
= =
= =

=


=


= =

,
1
1
,
2
1
2
2
1
2
2
1 1
2
13
Example 4.1

4.1.2 Model Adequacy Checking
Residual Analysis
Residual:
Basic residual plots indicate that normality,
constant variance assumptions are satisfied
No obvious problems with randomization


+ = = y y y y y y e
j i ij ij ij ij

14
DESIGN-EXPERT Pl ot
Hardness
Residual
N
o
r
m
a
l

%

p
r
o
b
a
b
i
l
i
t
y
Normal plot of residuals
-1 -0.375 0.25 0.875 1.5
1
5
10
20
30
50
70
80
90
95
99
15
DESIGN-EXPERT Pl ot
Hardness
2
2
2
2
Predicted
R
e
s
i
d
u
a
l
s
Residuals vs. Predicted
-1
-0.375
0.25
0.875
1.5
-2.75 -0.31 2.13 4.56 7.00
DESIGN-EXPERT Pl ot
Hardness
Run Number
R
e
s
i
d
u
a
l
s
Residuals vs. Run
-1
-0.375
0.25
0.875
1.5
1 4 7 10 13 16
16
Can also plot residuals versus the type of tip
(residuals by factor) and versus the blocks. Also
plot residuals v.s. the fitted values. Figure 4.5 and
4.6 in Page 137
These plots provide more information about the
constant variance assumption, possible outliers

4.1.3 Some Other Aspects of the Randomized
Complete Block Design
The model for RCBD is complete additive.
1, 2,...,
1, 2,...,
ij i j ij
i a
y
j b
t | c
=

= + + +

=

17
Interactions?
For example:


The treatments and blocks are random.
Choice of sample size:
Number of blocks |, the number of replicates
and the number of error degrees of freedom |

j i ij j i ij
y E y E | t | t ln ln ln ) ( ln ) ( + + = =
2
1
2
2
o
t
a
b
a
i
i
=
= u
18
Estimating miss values:
Approximate analysis: estimate the missing
values and then do ANOVA.
Assume the missing value is x. Minimize SS
E
to
find x

Table 4.8
Exact analysis


) 1 )( 1 (
' ' '

+
=

b a
y by ay
x
j i
19
4.1.4 Estimating Model Parameters and the General
Regression Significance Test
The linear statistical model


The normal equations
1, 2,...,
1, 2,...,
ij i j ij
i a
y
j b
t | c
=

= + + +

=

b b a
a
a b a
b
b a
y a a
y a a
y b b
y b b
y a a b b ab

= + + + +
= + + + +
= + + + +
= + + + +
= + + + + + +
| t t
| t t
| | t
| | t
| | t t

1
1 1 1
1
1 1 1
1 1

20
Under the constraints,

the solution is

and the fitted values,

The sum of squares for fitting the full model:


The error sum of squares
0

, 0
1 1
= =

= =
b
j
j
a
i
i
| t

= = = y y y y y
j j i i
| t

, ,

+ = + + = y y y y
j i j i ij
| t


ab
y
a
y
b
y
y y y R
b
j
j
a
i
i
b
j
j j
a
i
i i
2
1
2
1
2
1 1

) , , (

=

=

=

+ = + + =

| t | t
( )
2
1 1 1 1
2
) , , (

= =

= =
+ = =
a
i
b
j
j i ij
a
i
b
j
ij E
y y y y R y SS | t
21
The sum of squares due to treatments:
ab
y
b
y
R
a
i
i
2
1
2
) , | (

=

| t
22
4.2 The Latin Square Design
RCBD removes a known and controllable
nuisance variable.
Example: the effects of five different formulations
of a rocket propellant used in aircrew escape
systems on the observed burning rate.
Remove two nuisance factors: batches of raw
material and operators
Latin square design: rows and columns are
orthogonal to treatments.
23
The Latin square design is used to eliminate two
nuisance sources, and allows blocking in two
directions (rows and columns)
Usually Latin Square is a p p squares, and each
cell contains one of the p letters that corresponds
to the treatments, and each letter occurs once and
only once in each row and column.
See Page 145



24
The statistical (effects) model is



y
ijk
is the observation in the ith row and kth
column for the jth treatment, is the overall
mean, o
i
is the ith row effect, t
j
is the jth
treatment effect, |
k
is the kth column effect and
c
ijk
is the random error.
This model is completely additive.
Only two of three subscripts are needed to
denote a particular observation.
1, 2,...,
1, 2,...,
1, 2,...,
ijk i j k ijk
i p
y j p
k p
o t | c
=

= + + + + =

25
Sum of squares:
SS
T
= SS
Rows
+ SS
Columns
+ SS
Treatments
+ SS
E

The degrees of freedom:
p
2
1 = p 1 + p 1 + p 1 + (p 2)(p 1)
The appropriate statistic for testing for no
differences in treatment means is


ANOVA table (Table 4-10) (Page 146)
Example 4.3
) 1 )( 2 ( , 1 0
~

=
p p p
E
Treatments
F
MS
MS
F
26
The residuals

Table 4.13
If one observation is missing,


Replication of Latin Squares:
Three different cases
See Table 4.14, 4.15 and 4.16
Crossover design: Pages 150 and 151

+ = = y y y y y y y e
k j i ijk ijk ijk ijk
2

) 1 )( 2 (
2 ) (
' ' ' '

+ +
=

p p
y y y y p
y
k j i
ijk
27
4.3 The Graeco-Latin Square Design
Graeco-Latin square:
Two Latin Squares
One is Greek letter and the other is Latin letter.
Two Latin Squares are orthogonal
Table 4.18
Block in three directions
Four factors (row, column, Latin letter and
Greek letter)
Each factor has p levels. Total p
2
runs
28
The statistical model:

y
ijkl
is the observation in the ith row and lth
column for Latin letter j, and Greek letter k
is the overall mean, u
i
is the ith row effect, t
j

is the effect of Latin letter treatment j , e
k
is the
effect of Greek letter treatment k,
l
is the
effect of column l.
ANOVA table (Table 4.19)
Under H
0
, the testing statistic is F
p-1,(p-3)(p-1)

distribution.
Example 4.4
p l k j i y
ijkl l k j i ijkl
, , 1 , , , , = + + + + + + = c e t u
29
4.4 Balance Incomplete Block
Designs
May not run all the treatment combinations in
each block.
Randomized incomplete block design (BIBD)
Any two treatments appear together an equal
number of times.
There are a treatments and each block can hold
exactly k (k < a) treatments.
For example: A chemical process is a function of
the type of catalyst employed. See Table 4.22
30
4.4.1 Statistical Analysis of the BIBD
a treatments and b blocks. Each block contains k
treatments, and each treatment occurs r times.
There are N = ar = bk total observations. The
number of times each pairs of treatments appears
in the same block is

The statistical model for the BIBD is


) 1 (
) 1 (

=
a
k r

ij j i ij
y c | t + + + =
31
The sum of squares









Blocks adjusted Treatments T E
b
j
j ij i i
a
i
i
adjusted Treatments
b
j
j Blocks
i j
ij T
E Blocks Treatments T
SS SS SS SS
y n
k
y Q
a
Q k
SS
N y y
k
SS
N y y SS
SS SS SS SS
=
= =
=
=
+ + =

=

=

=


) (
1
1
2
) (
2
1
2
2 2
1
,
/
1
/

32
The degree of freedom:
Treatments(adjusted): a 1
Error: N a b 1
The testing statistic for testing equality of the
treatment effects:


ANOVA table (see Table 4.23)
Example 4.5
E
adjusted Treatments
MS
MS
F
) (
0
=
33
4.4.2 Least Squares Estimation of the Parameters
The least squares normal equations:






Under the constrains,

we have
j j
a
i
i ij j
i
b
j
j ij i i
b
j
j
a
i
i
y k n k
y n r r
y k r N

=

=

= =
= + +
= + +
= + +

| t |
| t t
| t

:
1
1
1 1
0

, 0
1 1
= =

= =
b
j
j
a
i
i
| t

= y
34
For the treatment effects,

a i
a
kQ
kQ k r
i
i
i
a
i p p
p i
, , 2 , 1 ,
) 1 (
, 1
= =
=

= =

t
t t