Chapter 6

Chapter 6

Balanced Incomplete Block Design (BIBD)

The designs like CRD and RBD are the complete block designs. We now discuss the balanced incomplete block design (BIBD) and the partially balanced incomplete block design (PBIBD) which are the incomplete block designs.

A balanced incomplete block design (BIBD) is an incomplete block design in which

· b blocks have the same number k of plots each and

· every treatment is replicated r times in the design.

·

Each treatment occurs at most once in a block, i.e., or 1 where is the number of times the jth treatment occurs in ith block,

·

Every pair of treatments occurs together is of the b blocks.

Five parameters denote such design as .

The parameters are not chosen arbitrarily.

They satisfy the following relations:

and for all Obviously cannot be a constant for all j. So the design is not orthogonal.

Example of BIBD

Now we see how the conditions of BIBD are satisfied.

Even if the parameters satisfy the relations, it is not always possible to arrange the treatments in blocks to get the corresponding design.

The necessary and sufficient conditions to be satisfied by the parameters for the existence of a BIBD are not known.

The conditions (I)-(III) are some necessary condition only. The construction of such design depends on the actual arrangement of the treatments into blocks and this problem is handled in combinatorial mathematics. Tables are available, giving all the designs involving at most 20 replication and their method of construction.

Theorem:

Proof: (I)

Let the incidence matrix

Observing that the quantities are the scalars and the transpose of each other, we find their values.

Consider

Similarly,

Proof: (II)

Consider

Since or 0 as or 0,

Also

Proof: (III)

From (I), the determinant of is

because since if from (II) that k = v. This contradicts the incompleteness of the design.

Thus is a nonsingular matrix.

Thus

We know from matrix theory result

But , there being b rows in N.

Thus

Interpretation of conditions of BIBD

Interpretation of (I) bk = vr

This condition is related to the total number of plots in an experiment. In our settings, there are plots in each block and there are blocks. So the total number of plots are .

Further, there are treatments and each treatment is replicated times such that each treatment occurs atmost in one block. So total number of plots containing all the treatments is . Since both the statements counts the total number of plots, hence

Interpretation of (II)

Each block has plots. Thus the total pairs of plots in a block =

There are blocks. Thus the total pairs of plots such that each pair consists of plots within a block = .

There are treatments, thus the total number of pairs of treatment = .

Each pair of treatment is replicated times, i.e., each pair of treatment occurs in blocks.

Thus the total number of pairs of plots within blocks must be .

Hence

Using in this relation, we get

Proof of (III) was given by Fisher but quite long, so not needed here.

Balancing in designs:

There are two types of balancing – Variance balanced and efficiency balanced. We discuss the variance balancing now and the efficiency balancing later.

Balanced Design (Variance Balanced):

A connected design is said to be balanced (variance balanced) if all the elementary contrasts of the treatment effects can be estimated with the same precision. This definition does not hold for the disconnected design, as all the elementary contrasts are not estimable in this design.

Proper Design:

An incomplete block design with is called a proper design.

Symmetric BIBD:

A BIBD is called symmetrical if the number of blocks = number of treatments, i.e.,

Since so from

Thus the number of pairs of treatments common between any two blocks = .

The determinant of is

When BIBD is symmetric, b = v and then using we have Thus

Since is an integer, hence when is an even number, must be a perfect square. So

Post-multiplying both sides by , we get

Hence in the case of a symmetric BIBD, any two blocks have treatment in common.

Since BIBD is an incomplete block design. So every pair of treatment can occur at most once is a block, we must have

If then it means that each treatment occurs once in every block which occurs in case of RBD.

So in BIBD, always assume v > k .

Similarly

Resolvable design:

A block design of

· b blocks in which

· each of v treatments is replicated r times

is said to be resolvable if b blocks can be divided into r sets of blocks each, such that every treatment appears in each set precisely once. Obviously, in a resolvable design, b is a multiple of r.

Theorem: If in a BIBD is divisible by r, then

Proof: Let (where is a positive integer).

Now, if possible, let

Intrablock analysis of BIBD:

Consider the model

where

We don’t need to develop the analysis of BIBD from starting. Since BIBD is also an incomplete block design and the analysis of incomplete block design has already been presented in the earlier module, so we implement those derived expressions directly under the setup and conditions of BIBD. Using the same notations, we represent the blocks totals by , treatment totals by , adjusted treatment totals by and grand total by The normal equations are obtained by differentiating the error sum of squares. Then the block effects are eliminated from the normal equations and the normal equations are solved for the treatment effects. The resulting intrablock equations of treatment effects in matrix notations are expressible as

.

in the case of BIBD. The diagonal elements of are given by

The off-diagonal elements of are given by

The adjusted treatment totals are obtained as

where denotes the sum over those blocks containing jth treatment. Denote

The matrix is simplified as follows:

Since is not as a full rank matrix, so its unique inverse does not exist. The generalized inverse of is denoted as which is obtained as

Since

the generalized inverse of is

Thus .

Thus an estimate of is obtained from as

The null hypothesis of our interest is against the alternative hypothesis at least one pair of is different. Now we obtain the various sum of squares involved in the development of analysis of variance as follows.

The adjusted treatment sum of squares is

The unadjusted block sum of squares is

.

The total sum of squares is

The residual sum of squares is obtained by

.

A test for is then based on the statistic

If

This completes the analysis of variance test and is termed as intrablock analysis of variance. This analysis can be compiled into the intrablock analysis of variance table for testing the significance of the treatment effect given as follows.

Intrablock analysis of variance table of BIBD for

Source

Sum of squares

Degrees of

freedom

Mean squares

F

Between treatment (adjusted)

Between blocks

(unadjusted)

Intrablock error

b - 1

bk – b – v + 1

Total

In case, the null hyperthesis is rejected, then we go for a pairwise comparison of the treatments. For that, we need an expression for the variance of the difference of two treatment effects.

The variance of an elementary contrast under the intrablock analysis is

This expression depends on which is unknown. So it is unfit for use in the real data applications. One solution is to estimate from the given data and use it in the place of

An unbiased estimator of is

Thus an unbiased estimator of can be obtained by substituting in it as

If is rejected, then we make pairwise comparison and use the multiple comparison test. To test a suitable statistic is

which follows a t-distribution with degrees of freedom under .

A question arises that how a BIBD compares to an RBD. Note that BIBD is an incomplete block design whereas RBD is a complete block design. This point should be kept in mind while making such restrictive comparison.

We now compare the efficiency of BIBD with a randomized block (complete) design with replicates. The variance of an elementary contrast under a randomized block design (RBD) is

where under RBD.

Thus the relative efficiency of BIBD relative to RBD is

The factor (say) is termed as the efficiency factor of BIBD and

The actual efficiency of BIBD over RBD not only depends on the efficiency factor but also on the ratio of variances . So BIBD can be more efficient than RBD as can be more than because

Efficiency balanced design:

A block design is said to be efficiency balanced if every contrast of the treatment effects is estimated through the design with the same efficiency factor.

If a block design satisfies any two of the following properties:

(i) efficiency balanced,

(ii) variance balanced and

(iii) an equal number of replications,

then the third property also holds true.

Missing observations in BIBD:

The intrablock estimate of missing (i, j)th observation is

the sum of Q value for all other treatment (but not the jth one) which are present in the

ith block.

All other procedures remain the same.

Interblock analysis and recovery of interblock information in BIBD

In the intrablock analysis of variance of an incomplete block design or BIBD, the treatment effects were estimated after eliminating the block effects from the normal equations. In a way, the block effects were assumed to be not marked enough and so they were eliminated. It is possible in many situations that the block effects are influential and marked. In such situations, the block totals may carry information about the treatment combinations also. This information can be used in estimating the treatment effects which may provide more efficient results. This is accomplished by an interblock analysis of BIBD and used further through the recovery of interblock information. So we first conduct the interblock analysis of BIBD. We do not derive the expressions a fresh but we use the assumptions and results from the interblock analysis of an incomplete block design. We additionally assume that the block effects are random with variance

After estimating the treatment effects under interblock analysis, we use the results for the pooled estimation and recovery of interblock information in a BIBD.

In case of BIBD,

The interblock estimate of can be obtained by substituting the expression on in the earlier obtained interblock estimate.

Our next objective is to use the intrablock and interblock estimates of treatment effects together to find an improved estimate of treatment effects.

In order to use the interblock and intrablock estimates of together through pooled estimate, we consider the interblock and intrablock estimates of the treatment contrast.

The intrablock estimate of treatment contrast is

The interblock estimate of treatment contrast is

The variance of is obtained as

Since

so

Similarly, the variance of is obtained as

The information on and can be used together to obtain a more efficient estimator of by considering the weighted arithmetic mean of and . This will be the minimum variance unbiased and estimator of when the weights of the corresponding estimates are chosen such that they are inversely proportional to the respective variances of the estimators. Thus the weights to be assigned to intrablock and interblock estimates are reciprocal to their variances as and respectively. Then the pooled mean of these two estimators is

where ,

Now we simplify the expression of so that it becomes more compatible in further analysis.

Since so the numerator of can be expressed as

Similarly, the denominator of can be expressed as

Let

where Using these results we have

where

Thus the pooled estimate of the contrast is

The variance of is

is called as the effective variance.

Note that the variance of any elementary contrast based on the pooled estimates of the treatment effects is

The effective variance can be approximately estimated by

where MSE is the mean square due to error obtained from the intrablock analysis as

and

.

The quantity depends upon the unknown and To obtain an estimate of , we can obtain the unbiased estimates of and and then substitute them back in place of and in . To do this, we proceed as follows.

An estimate of can be obtained by estimating from the intrablock analysis of variance as

.

The estimate of depends on . To obtain an unbiased estimator of , consider

for which

Thus an unbiased estimator of is

where

Thus

Recall that our main objective is to develop a test of hypothesis for and we now want to develop it using the information based on both interblock and intrablock analysis.

To test the hypothesis related to treatment effects based on the pooled estimate, we proceed as follows.

Consider the adjusted treatment totals based on the intrablock and the interblock estimates as

and use it as usual treatment total as in earlier cases.

The sum of squares due to is

Note that in the usual analysis of variance technique, the test statistic for such hull hypothesis is developed by taking the ratio of the sum of squares due to treatment divided by its degrees of freedom and the sum of squares due to error divided by its degrees of freedom. Following the same idea, we define the statistics

where is an estimator of . It may be noted that depends on The value of itself depends on the estimated variances . So it cannot be ascertained that the statistic necessary follow the distribution. Since the construction of is based on the earlier approaches where the statistic was found to follow the exact -distribution, so based on this idea, the distribution of can be considered to be approximately distributed. Thus the approximate distribution of is considered as distribution with and degrees of freedom. Also, is an estimator of which is obtained by substituting the unbiased estimators of and .

An approximate best pooled estimator of is

and its variance is approximately estimated by

In case of the resolvable BIBD, can be obtained by using the adjusted block with replications sum of squares from the intrablock analysis of variance. If sum of squares due to such block total is and corresponding mean square is

then

and for a resolvable design. Thus

and hence

The analysis of variance table for the recovery of interblock information in BIBD is described in the following table:

Source

Sum of squares

Degrees of freedom

Mean square

Between treatment

(unadjusted)

Between blocks

(adjusted)

Intrablock error

v - 1

b - 1

bk – b – v + 1

Total

bk - 1

The increase in the precision using interblock analysis as compared to intrablock analysis is

Such an increase may be estimated by

.

Although but this may not hold true for . The estimates may be negative also and in that case we take

Analysis of Variance | Chapter 6 | Balanced Incomplete Block Design | Shalabh, IIT Kanpur (23) (23)

k

(1)

.

2

2

k

kk

æö

-

=

ç÷

èø

b

(1)

2

kk

b

-

v

1,2,...,;1,2,...,.

ibjv

==

(1)

2

2

v

vv

æö

-

=

ç÷

èø

l

l

(1)

2

vv

l

-

=

(

)

1

(1)

22

vv

kk

b

l

-

-

=

bkvr

=

(

)

(

)

11.

rkv

l

-=-

12

....

b

kkkk

====

.

bv

=

,

bv

=

l

bkvr

=

.

kr

Þ=

l

'

NN

(

)

(

)

1

1

1

'[(1)]()

1

().

v

v

v

NNrvr

rrkr

rkr

ll

l

l

-

-

-

=+--

=+--

éù

ëû

=-

,

bkvr

=

.

kr

=

2

21

1

2

'(),

so

().

v

v

NNNrr

Nrr

l

l

-

-

==-

=±-

N

v

(,,,;)

Dbkvr

l

()

r

l

-

'

11

11'1

'

11

2

'1'

11

2

'(),

(')

1

,

1

.

vv

vv

vv

NNrIEE

NNNN

IEE

rr

NIEE

rr

ll

l

l

l

l

---

-

=-+

=

éù

=-

êú

-

ëû

éù

=-

êú

-

ëû

'

N

'

11

'()'.

vv

NNrIEENN

ll

=-+=

l

.

vk

³

,

vk

=

.

r

l

<

[Ifthen(1)(1) which means that the desig

n is RBD]

rvrkvk

ll

=-=-Þ=Þ

/

br

,,,and

bkvr

l

(,,,,),

Dvbrk

l

b

1.

bvr

³+-

bnr

=

1

n

>

ForaBIBD,(1)(1)

because

(1)

oror

(1)

or

(1)

(1)

1

.

1

Since1and1,so1isaninteger.Sincehastobean

integer.

(1)

isalsoapo

1

vrk

vrbk

v

rvrnrk

k

vnk

nk

k

n

n

k

nknr

n

k

l

l

l

ll

l

l

-=-

=

éù

-

êú

==

êú

-

êú

=

ëû

-

=

-

-

æö

=+

ç÷

-

èø

>>>

-

Þ

-

sitiveinteger.

1

1

or(1)1

(1)(1)

or(1)(because1)

(1)

1whichisacontradictionasintegercannotbel

essthanone

1

1isimpossible.Thustheoppositeistrue.

1holdscorrect.

bvr

nrvr

rnv

rkrk

rnv

n

k

bvr

bvr

ll

l

<+-

Þ<+-

-<-

--

-<-=

-

Þ<

-

Þ<+-

Þ³+-

;1,2,...,;1,2,...,,

ijijij

yibjv

mbte

=+++==

           is the general mean effect;

           is the fixed additive block

effect;

           is the fixed additive treatm

ent effect and

           is the i.i.d. random error wi

th ~

th

i

th

j

ijij

i

j

m

b

t

ee

2

(0,).

N

s

1

v

iij

j

By

=

=

å

()

() (1)(1)

() (andhence).

Ibkvr

IIvrk

IIIbvrk

l

=

-=-

³>

1

b

jij

i

Vy

=

=

å

j

Q

1

bv

ij

ij

Gy

==

=

åå

ˆ

QC

t

=

Nowweobtaintheformsofand

CQ

C

2

1

(1,2,...,)

.

b

ij

i

jj

n

crj

k

r

r

k

n

=

=-=

=-

å

C

''

1

1

(';,'1,2,...,)

.

b

jjijij

i

cnnjjjj

k

k

n

l

=

=-¹=

=-

å

1

()

1

(1,2,...,)

1

b

jjiji

i

ji

ij

QVnBj

k

VB

k

n

=

=-¹

=-

å

å

Hence forall

forall

ij

j

ij

j

nki

nrj

=

=

å

å

()

ij

å

()

,then

.

ji

ij

j

jj

TB

T

QV

k

=

=-

å

C

'

11

'

1

'

1

'

11

'

1

()

1

()

1

()

.

vv

vvi

vvi

vv

NN

CrI

k

rIrIEE

k

k

rIEE

kk

v

IEE

kk

EE

V

I

kv

ll

l

l

l

l

=-

éù

=--+

ëû

-

æö

=+-

ç÷

èø

-

æö

=+-

ç÷

èø

æö

=-

ç÷

èø

C

C

C

-

1

'

11

.

vv

EE

CC

v

-

-

æö

=+

ç÷

èø

'

11

'

11

or,

vv

v

vv

v

EE

v

CI

kv

EE

kC

I

vv

l

l

æö

=-

ç÷

èø

=-

k

C

v

l

'

1'2'

...

jj

jijjijbb

nnnnnn

l

+++=

1

1

'

11

1

''

1111

,

.

vv

vvvv

v

v

EE

k

CC

vv

EEEE

I

vv

I

l

-

-

-

-

éù

æö

=+

ç÷

êú

èø

ëû

éù

=-+

êú

ëû

=

v

v

CI

k

l

-

=

t

QC

t

=

ˆ

.

CQ

v

Q

k

t

l

-

=

=

012

:...

v

H

ttt

===

1

:

H

'

j

s

t

()

2

1

ˆ

'

'

,

Treatadj

v

j

j

SSQ

k

QQ

k

Q

t

ln

ln

=

=

=

=

å

2

2

()

1

b

i

Blockunadj

i

B

G

SS

kbk

=

=-

å

'1,2,...,.

jjv

¹=

2

2

11

bv

Totalij

ij

G

SSy

bk

==

=-

åå

()()()

ErrortTotalBlockunadjTreatadj

SSSSSSSS

=--

012

:...

v

H

ttt

===

()

()

2

1

()

/(1)

/(1)

1

..

1

Treatadj

Tr

Errort

v

j

j

Errort

SSv

F

SSbkbv

Q

kbkbv

vvSS

l

=

-

=

--+

--+

=

-

å

1,1,1;0()

thenisrejected.

Trvbkbvt

FFH

a

----+

>

012

:...

v

H

ttt

===

()

Treatadj

SS

()

Blockunadj

SS

()

(bysubstraction)

Errort

SS

1

v

-

ij

n

r

()

1

Treatadj

treat

SS

MS

v

=

-

()

1

Errort

E

SS

MS

bkbv

=

--+

Treat

E

MS

MS

2

2

Total

ij

ij

SS

G

y

bk

=

-

åå

1

bk

-

'

(,')

jj

jj

tt

-¹

'

'

2

'

22

2

2

'''

22

2

2

22

2

ˆˆ

*()

()

[()()2()]

(2)

12

21

2

.

jj

jj

jjjj

jjjjjj

VVar

k

VarQQ

v

k

VarQVarQCovQQ

k

ccc

k

r

kk

k

v

tt

l

ln

s

ln

l

s

ln

s

l

=-

æö

=-

ç÷

èø

=+-

=+-

éù

æö

=-+

ç÷

êú

èø

ëû

=

2

s

2

s

2

.

s

11016

Inthedesign(,;,;):consider10 (,,...,),

6 (,,...,), 3,5,2

DbkvrbsayBBvsayTTkr

ll

=====

2

σ

()

2

ˆ

.

1

Errort

SS

bkb

s

n

=

--+

*

V

2

ˆ

s

()

*

2

ˆ

..

1

Errort

SS

k

V

vbkb

ln

=

--+

0

H

0''

:('),

jj

Hjj

tt

=¹

'

()

(1)

.

jj

Errort

QQ

kbkbv

t

v

SS

l

-

--+

=

(1)

bkbv

--+

0

H

r

2

*2

*

'

2

ˆˆ

()

RjjRBD

VVar

r

s

tt

=-=

2

*

()

ij

Vary

s

=

2

*

'

2

'

2

*

2

2

ˆˆ

()

ˆˆ

()

2

.

jjRBD

jjBIBD

Var

r

Var

k

v

v

rk

s

tt

tt

s

l

s

l

s

æö

ç÷

-

èø

=

-

æö

ç÷

èø

æö

=

ç÷

èø

v

E

rk

l

=

1

1

1

11

11

1(since).

vvk

E

rkkv

kv

vk

l

-

-

æö

==

ç÷

-

èø

æöæö

=--

ç÷ç÷

èøèø

<>

22

*

/

ss

2

*

s

2

s

.

kv

<

ij

y

'

(1)(1)(1)

(1)(1)

ijj

ij

vrkBkvQvQ

y

kkbkbv

-----

=

---+

'

:

j

Q

2

.

b

s

2

1121

2

1222

2

12

'

11

'

1

11

'

()

1

(')

iiiiiv

iii

iiiiiv

iii

iviiviiv

iii

vvv

vv

v

nnnnn

nnnnn

NN

nnnnn

r

r

r

rIEE

EE

NNI

rrk

ll

ll

ll

ll

l

l

-

æö

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

æö

ç÷

ç÷

=

ç÷

ç÷

èø

=-+

éù

=-

êú

-

ëû

ååå

ååå

ååå

L

L

MMOM

L

L

L

MMOM

L

123

124

134

146

156

236

245

256

345

346

1

2

3

4

Blocks  Treatm

5

6

7

8

9

1

ents

0

BTTT

BTTT

BTTT

BTTT

BTTT

BTTT

BTTT

BTTT

BTTT

BTTT

t

(

)

1

'

NN

-

1

1

(')'.

v

GE

NNNB

bk

t

-

=-

%

t

'

l

t

ˆ

''

'

ˆ

,say.

jj

j

jj

j

llCQ

k

lQ

v

k

lQ

v

l

t

l

l

t

-

=

=

=

=

å

å

'

l

t

1

11

1

1

''

'(since'0)

1

1

.

v

vb

jiji

ji

v

jj

j

v

jj

j

lNB

llE

r

lnB

r

lT

r

l

t

l

l

l

t

==

=

=

==

-

æö

=

ç÷

-

èø

=

-

=

åå

å

å

%

%

ˆ

'

l

t

2

2

''

'()

ˆ

(')

()2(,).

jj

j

jjjjjj

jjjj

k

VarlVarlQ

v

k

lVarQllCovQQ

v

t

l

l

¹

æö

æö

=

ç÷

ç÷

èø

èø

éù

æö

=+

êú

ç÷

èø

ëû

å

ååå

() 10330and6530

() (1)2510and(1)5210

(1)(1)

() 106

ibkvr

bkvr

iivrk

vrk

iiib

l

l

=´==´=

Þ=

-=´=-=´=

Þ-=-

=³

2

2

'

1

()1,

(,),('),

j

jj

VarQr

k

CovQQjj

k

s

l

s

æö

=-

ç÷

èø

=-¹

[

]

2

2

2222

2

222

j

j

2

2

1

ˆ

(')1

(1)

(since 0beingcontrast)

1

(1)(using(1)(1))

jjj

jjj

jj

jj

j

j

k

Varlrlll

vkk

krk

ll

vkk

k

vlrkv

vk

k

v

l

tss

l

l

s

l

lll

l

l

éù

ìü

æö

ïï

æöæö

êú

=---

íý

ç÷

ç÷ç÷

êú

èøèø

èø

ïï

îþ

ëû

éù

-

æö

=+=

êú

ç÷

èø

ëû

æö

=-+-=-

ç÷

èø

æö

=

ç÷

èø

ååå

ååå

å

l

22

.

j

j

l

s

å

ˆ

'

t

l

2

2

''

'()

2

2

2222

2

2

1

(')()2(,)

1

.

jjjjjj

jjjj

fjfjj

jjj

f

j

j

VarllVarTllCovTT

r

rlll

r

l

r

t

l

sls

l

s

l

¹

éù

æö

=+

êú

ç÷

-

èø

ëû

éù

ìü

æö

ïï

æö

êú

=+-

íý

ç÷

ç÷

-

êú

èø

èø

ïï

îþ

ëû

=

-

ååå

ååå

å

%

ˆ

'

t

l

ˆ

'

t

l

'

t

l

ˆ

'

t

l

'

t

%

l

'

t

l

()

()(1)(1)

().

Ibkvr

IIvrk

IIIbv

l

=

-=-

³

2

/()

vk

ls

2

()/,

f

r

ls

-

2222

2222

1

2

12

12

12

12

12

ˆˆˆ

''

*

ˆ

()

()

ˆ

()

()

ˆ

()

()

jjjj

jj

ff

ff

jjjj

jj

jjjj

jj

jj

j

j

vrvr

llll

kk

L

vrvr

kk

v

lrl

k

v

r

k

vlkrl

vkr

vkr

l

vkr

llll

tttt

ssss

llll

ssss

lw

tlwt

l

wlw

lwtlwt

lwlw

lwtlwt

lwlw

--

++

==

--

++

+-

=

+-

+-

=

+-

+-

éù

=

êú

+-

ëû

=

åå

åå

åå

å

%

%

%

%

*

jj

j

l

t

å

12

*

12

ˆˆ

()

()

jj

j

kr

kr

lnwtlwt

t

lnwlw

+-

=

+-

12

22

11

,.

f

ww

ss

==

*

j

t

ˆ

(/)and/(),

jjjj

kQTr

tlntl

==-

%

*

j

t

1212

ˆ

()

jjjj

krkQkT

wlntwltww

+-=+

%

*

j

t

():

ij

Nnbv

=´

[

]

12

12

12

()

(1)(1)

(using(1)(1))

11

1

(1)().

1

vkr

vrkrk

krvrk

vv

vrkkrvk

v

wlwl

wwl

ww

+-

-é-ù

éùæö

=+--=-

ç÷

êú

êú

--

ëûèø

ëû

=-+-

-

*

()(1)(1)

jjj

WvkVvTkG

=---+-

*

0.

j

j

W

=

å

[

]

[

]

12

*

12

12

12

121

12

*

112

(1)

(1)()

(1)()

= (using)

(1)()

(1)()(1)

=

(1)()

(1)()()(1)

jj

j

jjj

j

jj

jj

jjj

vkQkT

rvkkrvk

vkVTkT

T

QV

rvkkvkk

kvVkvT

rvkkvk

kvVkWvkVkG

ww

t

ww

ww

ww

www

ww

www

éù

-+

ëû

=

-+-

éù

--+

ëû

=-

-+-

-+--

-+-

é

-+-----

ë

=

[

]

[

]

[

]

{

}

{

}

12

*

11212

12

*

12

12

*

(1)()

(1)()()()(1)

(1)()

1

(1)

(1)()

1

(1)

jj

jj

jj

rvkkvk

kvkvkVkWkG

rvkkvk

k

VWkG

rvkkvk

VWkG

r

ww

wwwww

ww

ww

ww

x

ù

û

-+-

éù

----+---

ëû

=

-+-

éù

-

=+--

êú

-+-

ëû

éù

=+--

ëû

12

12

22

12

11

,,.

(1)()

f

k

vkkvk

ww

xww

wwss

-

===

-+-

'

l

t

*

*

'*

1

()(since0beingcontrast)

jj

j

jjjj

jj

ll

lVWl

r

tt

x

=

=+=

å

åå

'*

l

t

[

]

2

12

2

12

2

2

2

12

('*)

()

(1)

(using(1)(1)

(1()

where

(1)

(1)()

j

j

j

j

j

j

E

E

k

Varll

vkr

kv

lvrk

rvkkvk

l

r

kv

vkkvk

t

lwlw

l

ww

s

s

ww

=

+-

-

=-=-

-+-

=

-

=

-+-

å

å

å

**2

2

().

ijE

Var

r

tts

-=

1111

and'

bvvb

ENEENE

[

]

2

ˆ

1()*

E

MSEvk

sw

=+-

()

1

Errort

SS

MSE

bkbv

=

--+

12

12

*

(1)()

vkkvk

ww

w

ww

-

=

-+-

*

w

2

s

2

.

b

s

*

w

2

s

2

b

s

2

s

2

b

s

*

w

1

w

2

s

1

1

2

1

ˆ

[]

ˆ

MSE

w

s

-

==

2

w

22

ˆˆ

and

b

ss

2

b

s

()()()()

BlockadjTreatadjBlockunadjTreatunadj

SSSSSSSS

=+-

22

()

()()(1).

Blockadj

ESSbkvb

b

ss

=-+-

11211

12222

11

12

1

2

1

1

1

(1,1,...,1)

1

(1,1,...,1)

(1,1,...,1)= .

b

b

bv

vvbv

j

j

j

j

bj

j

b

nnn

nnn

ENE

nnn

n

n

n

k

k

bk

k

´

æö

æö

ç÷

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

èø

æö

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

æö

ç÷

ç÷

=

ç÷

ç÷

èø

å

å

å

L

L

M

MMOM

L

M

M

2

b

s

22

()

()

()

()

1

ˆˆ

(1)

1

(1)

1

1

(1)

Blockadj

Blockadj

Blockadj

Blockadj

SSb

bkv

SSbMSE

bkv

b

MSMSE

bkv

b

MSMSE

vr

b

ss

éù

=--

ëû

-

éù

=--

ëû

-

-

éù

=-

ëû

-

-

éù

=-

ëû

-

()

()

.

1

Blockadj

Blockadj

SS

MS

b

=

-

2

22

()()

1

ˆ

ˆˆ

1

.

(1)(1)()

BlockadjErrort

k

vrkbSSvkSS

b

w

ss

=

+

=

éù

----

ëû

012

:...

v

H

ttt

===

**

*;1,2,...,

jjj

TTWjv

w

=+=

*

j

T

2

*

1

2*2

*

1

.

v

j

v

j

Tj

j

T

ST

v

=

=

æö

ç÷

èø

=-

å

å

2

*

*

/[(1)]

ˆ

[1()*]

T

Svr

F

MSEvk

w

-

=

+-

ˆ

*

w

11211

12222

11

12

1

1

'(1,...,1)

1

=(1,1,...,1)(1,1,...,1)

b

b

vb

vvbv

i

i

v

iv

i

nnn

nnn

ENE

nnn

n

r

vr

r

n

´

æö

æö

ç÷

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

èø

æö

æö

ç÷

ç÷

ç÷

==

ç÷

ç÷

ç÷

ç÷

èø

ç÷

èø

å

å

L

L

M

M

MMOM

L

MM

*

w

*

F

ˆ

*.

w

ˆ

*

w

22

ˆˆ

and

f

ss

*

F

F

*

F

F

*

F

1111

But

'asbotharescalars.

Thus .

bvvb

ENEENE

bkvr

=

=

F

*

F

F

(1)

v

-

(1)

bkbv

--+

ˆ

*

w

*

w

1

w

2

w

1

jj

j

l

t

=

å

1121111121

1222221222

1212

2

1121

2

1222

2

12

'

bv

bv

vvbvbbbv

iiiiiv

iii

iiiiiv

iii

iviiviiv

iii

nnnnnn

nnnnnn

NN

nnnnnn

nnnnn

nnnnn

nnnnn

æöæö

ç÷ç÷

ç÷ç÷

=

ç÷ç÷

ç÷ç÷

èøèø

=

ååå

ååå

åå

LL

LL

MMOMMMOM

LL

L

L

MMOM

L

.(1)

r

r

r

ll

ll

ll

æö

ç÷

ç÷

ç÷

ç÷

ç÷

ç÷

ç÷

èø

æö

ç÷

ç÷

=

ç÷

ç÷

èø

å

L

L

MMM

L

1

ˆ

v

jj

j

j

VW

l

r

x

=

+

å

2

12

.

ˆˆ

()

j

j

kl

vrk

lwlw

+-

å

2

ˆ

b

s

*

Block

SS

*

*

Block

Block

SS

MS

br

=

-

*22

22

()(1)

()

(1)

Block

vkr

EMS

br

rk

r

b

b

ss

ss

--

=+

-

-

=+

()()

kbrrvk

-=-

*22

(1)()

Block

ErMSMSErk

b

ss

éù

-=-+

ëû

[

]

1

*

2

1

1

ˆ

,

1

ˆ

.

block

rMSMSE

r

MSE

w

w

-

-

éù

-

=

êú

-

ëû

=

*

F

2

1

ij

n

=

2

*

T

S

()

()

()

()

Blockadj

Treatadj

Blockunadj

Treatunadj

SS

SS

SS

SS

-

=

+

()

(

bysubstraction)

Errort

SS

()

()

1

Blockadj

Blocksadj

SS

MS

b

=

-

()

1

Errort

SS

MSE

bkbv

=

--+

()

*

Blocksadj

MS

F

MSE

=

Total

SS

12

1

2

1

ˆ

()

1

(*)

()

1

()

.

Var

Var

vkr

v

rk

v

t

t

lwwl

lw

wl

lw

-

+-

=-

-

=

2

1

ˆ

()

ˆ

rk

v

wl

lw

-

12

ww

>

1

ij

n

=

12

ˆˆ

and

ww

12

ˆˆ

and

ww

12

ˆˆ

.

ww

=

2

soNumber

=

of

for all 1,2,...,oftimesoccursinthedesi

times   occurs in the design

gn

ijj

i

n

rjv

t

=

=

å

''

of blocks in which   occurs together

and Number and

= forall'.

ijijjj

i

nn

jj

tt

l

=

¹

å

1

1

1

1

'

1

(1)

(1)

[(1)].(2)

(1)

v

v

r

r

NNE

r

rv

rv

rvE

rv

ll

ll

ll

l

l

l

l

æöæö

ç÷ç÷

ç÷ç÷

=

ç÷ç÷

ç÷ç÷

èøèø

+-

æö

ç÷

+-

ç÷

==+-

ç÷

ç÷

+-

èø

L

L

MMOMM

L

M

11121

21222

1

12

1

2

11211

12222

2

1

1

1

''

1

'

v

v

v

bbbv

j

j

j

j

bj

j

b

b

ivvbv

b

nnn

nnn

NNEN

nnn

n

n

N

n

nnn

k

nnn

k

k

nnn

´

éù

æö

æö

êú

ç÷

ç÷

êú

ç÷

ç÷

=

êú

ç÷

ç÷

êú

ç÷

ç÷

êú

èø

èø

ëû

æö

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

æö

æö

ç÷

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

èø

å

å

å

L

L

M

MMOM

L

M

L

L

M

M

L

1

2

1

(3)

i

i

i

i

iv

i

v

n

n

k

n

r

r

k

r

krE

æö

ç÷

ç÷

ç÷

=

ç÷

ç÷

ç÷

ç÷

èø

æö

ç÷

ç÷

=

ç÷

ç÷

èø

=

å

å

å

M

M

(

)

(

)

11

[(1)]

or (1)

or (1)(1)

From 2 and 3

vv

rvEkrE

rvkr

vrk

l

l

l

+-=

+-=

-=-

'

NN

0

ij

n

=

[

]

(

)

1

1

1

det'[(1)]()

(1)

()

0

v

v

v

NNrvr

rrkr

rkr

ll

l

l

-

-

-

=+--

=+--

=-

¹

r

l

=Þ

'

NN

vv

´

(').

rankNNv

=

()(')

so ()

rankNrankNN

rankNv

=

=

()

rankNb

£

.

vb

£

k

b

ij

n

bk

v

r

vr

.

bkvr

=