An Assignment on Nested Design

Course No: Stat-3105

Course Title: Experimental Design

Submitted to:
Sharlene Alauddin
Lecturer Submitted by:
Statistics Discipline S.M. Khalid
Syfullah
Khulna University ID:172011
Khulna. Statistics
Discipline,
Khulna
University,
Khulna.

Date Of Submission:4th August , 2020.

Question No- 1 : Explain the concept of Nested design. Why it
is called balanced Nested design. Describe the situation where
this type of design is appropriate.
Answer:
Nested Designs : A nested design (sometimes referred to as a
hierarchical design) is used for experiments in which there is an
interest in a set of treatments and the experimental units are sub-
sampled.
Nested design is a analyze design in which levels of one factor (
say, Factor B) are hierarchically subsumed under ( or nested
within) levels of another factor ( say, Factor A).As a result,
assessing the complete combination of A and B levels is not
possible in a nested design.
Example: Consider a typical provenance study where garments
workers collects 5 cloths from 5 good color in each of 3 industries.
The cloth are produced in garments and the color are measured
for brightness of the cloth.

Balanced Nested design: In two-stage Nested design for

factor A and B, every level of factor B does not occur with every
level of factor A, there is no interaction effect between A and B.
This is a balanced nested design as B maintains equal number of
levels for each level of A and equal number of replicates prevails
under each level of combination. For this reason it is called
balanced nested design.
The situation where Nested design is appropriate :
Nested models are used for several statistical tests and analyses,
including multiple regression , likelihood – ration tests , conjoint
analysis , and independent of irrelevant alternative (IIA) . While
the above definition gets a bit more technical depending on
where you are using is.
For example,
In multiple regression and structural equation modelling
(SEM) , the idea is the same – that one model is nested inside
another . More technically , both models must have identical
terms and one of the models must have one or more extra
terms.
For example:
Y=0+11+22
Y= 0+ 11+22+ 312

The large model is called the full model and the smaller model is
called the reduced model.

Question-2: Consider a typical provenance study where a forest

geneticist collects 5 seeds from 5 superior trees in each of 3
forests. The seeds are germinated in a greenhouse and the
seedlings are measured for height growth. Test that whether
there is any significant difference among the effects of the factors.
Tree Forest
A B C D E
15.8 18.5 12.3 19.5 16.0
1
15.6 18.0 13.0 17.5 15.7
16.0 18.4 12.7 19.1 16.1
1 47.4 54.9 38.0 56.1 47.8
13.9 17.9 14.0 18.7 15.8
2
14.2 18.1 13.1 19.0 15.6
13.5 17.4 13.5 18.8 16.3
2 41.6 54.4 40.6 56.5 47.7
484.0
89.0 108.3 78.6 112.6 95.5

Answer :
For computing,
the Nested design the ANOVA table is:
 Table

Sum of Df SS MS F
Squares
Among p-1 SSa MSa MSa/MSb
Forests
Btwn trees P(q-1) SSb MSb MSb/MSe
within
forests
Amng Pq(r-1) SSe MSe
seedling
within trees
Total Pqr-1

Here, the value of G is 484

2
Total SS = ∑ =1 ∑ =1 ∑ =1 2-

=7949.22 -7808.53
=140.69

..2 2

SS(A) =∑ -
=1
= 7937.81 - 7808.53
 =129.8
∑ ∑ 2 2
SS(B) = .-
=1 =1
 =7945.21 – 7937.81
=7.4
SS(W) =Total SS – SS(A) – SS(B)
= 140.69 – 129.8 – 7.4
= 4.01

ANOVA
Table

Source of DF SS MS F
variation
Among 4 129.28 32.32 22.6
Forests
Trees 5 7.14 1.43 7.15
(Forest)

Among 20 4.01 0.20

Seedlings
Total 29
In the Nested design 5 forests, 2 trees per forest (10 total), 3
seedlings grown from each tree. Seedlings are nested within
tree are nested within forest.
Hypothesis:
H0 = There are no significant differences among the
factor H1: There are significant differences among the
factor

In the Nested design 5 forests, 2 trees per forest (10 total), 3

seedlings grown from each tree. Seedlings are nested within
tree are nested within forest.
Difference in coding Nested design, each tree must be coded
differently as one tree cannot occur in five different forests.
Using R software we can calculate the significant difference
among the effects of the factors.
Comment: In the ANOVA table we can see that
the p value is 0.05. We know that if the p value is ≤
0.05 then it indicates strong evidence against the
null hypothesis. So the null hypothesis (H0) is
rejected and H1 is accepted. Since the p vale is 0.05,
so we can say there is any significant difference
among the effects of the factors.

Question – 03: Discuss how you can data of a balanced

two stage Nested design.
Answer :
Random effect model with A fixed B random is also frequently
encountered. Fixed effects model for two – stage nested design is
given by
The two-stage nested model is:
= μ+ + ( )+ ( )

Where µ is the overall mean, αi is the ith factor A effect, βj(i) is

the jth effect of factor B nested within the ith level of factor A, is

the random error of the kth observation from the jth level of B
within

the ith level of A.

If we impose the constraints,

∑ =1 = 0
∑ =1 ( ) =0 for i=1,2,…………….,a

then the least squares estimates of the

model parameters are
̂= = ̂ ( )
̂ =

If we substitute these estimates, we get

= μ+ + ( )+ ( )
=y··· + (yi·· − y···) + (yij· − yi··) + eijk
 where eijk is the kth residual from the (i,j)th nested .
treatment.

Thus eijk = is the error

Notation for an ANOVA

= the sum of squares for A (df

= a − 1)
• MSA = SSA/(a − 1) = the mean square for A

= the sum of squares for B nested

within A (df = a(b − 1))
• MSB(A) = SSB/[a(b − 1)] = the mean square for B nested
within A
= the error sum of squares (df =
ab(n − 1))
• MSE = SSE/ab(n − 1)= the mean square error

= the total sum of squares (df

= abn − 1)
• Like previous designs, the total sum of squares for the two
factor CRD is partitioned into components corresponding to
the terms in the model:

OR
SST = SSA + SSB(A) + SSE
 • The alternate SS formulas for the balanced two stage nested
design are:

SSE = SST − SSA − SSB(A)

ANOVA Table for Two-Stage Balance Nested Design :

Source of Sum of Mean F
Variatio Square
n s d.f. Square Ratio

FA = (see ‡
MSA = SSA/(a
A SSA a−1 − 1) below)
MSB = SSB(A)/[a(b − FB =
B(A) SSB(A) a(b -1) 1)] MSB(A)/MSE
Error SSE ab(n - MSE = SSE/[ab(n − ——
1) 1)]

abn −
Total SStotal 1 —— ——
‡If B(A) is a fixedfactor then FA = MSA/MSE
If B(A) is a random factor then FA = MSA/MSB(A)

• To estimate variance components, we use the same approach

that was used for the one- and two-factor random effects
models:
If A and B(A) are random, replace E(MSA), E(MSB((A)), and
E(MSE) in
the expected means square equations with the calculated
values of
MSA, MSB(A), and MSE.
 • Solving the system of equations produces estimates of the
variance components:

σb2 = MSE

Question – 04: Set up liner model of this design (balanced

two stage Nested design) and hence estimate the model
parameters.

Answer :
Linear model for balanced two stage Nested design:
∴ = + + ( )+ ( )
Here, = General mean
= Fixed effect of ith level of A

( )= Fixed effect of jth level of B within ith

level of A ( )= Random error term
Assumption:

1. , αi and βj(i) are unknown

parameters. =0 and
3. εk(ij)~ NID(0,σ2)
Now error SS,

E=∑ =1 ∑ =1 ∑ =1 ( ) 2
=∑ ∑ ∑ ( yijk − μ − αi − βj(i))2
=1 =1 =1

Now,

( yijk − μ − αi − βj(i)) = 0

= −2 ∑ ∑∑ (yijk − μ − αi − βj(i)) = 0
=1 =1 =1

= - 2∑ =1(y − μ − α − βj(i))
( )
∑ ∑ =1 ∑ =1 yijk = pqrμ + qr ∑ =1 αi + r ∑ =1 ∑ =1 βj(i) … (1)
=1
∑ =1 ∑ =1 = qrμ + qrαi + r ∑ βj(i) … … (2)
⇒ ∑ yijk = rμ + rαi + rβj(i) … … (3)
These equations are known as normal equations.
Now imposing the restriction ∑pi=1 αi = 0 and ∑qj=1 βj(i) = 0
We get, from equation (1)
∑ ∑ ∑
=1
∴ ̂= =1 =1 = ̅

From equation (2) we get,

qru + qr = ∑ =1 ∑ =1
∑ ∑
̂= =1 =1 - ̂
=̅̅ - ̅
..
From equation (3) we get,
∴ ( ) = ∑ =1 - ̂- ̂
=̅- ̅-̅+ ̅
. ..
= ̅̅ - ̅
. ..

Thus we have obtained the least square estimates of the

parameters of the model .
In random effects models, we get
̂
2
2
=
2 2
̂=

2 −

2 2
̂=
2 −
∴ This is the estimate value for all estimator .