Chapter-6

Nested and Split Plot Designs

Both designs have reasonably widespread
application in the industrial use of designed
experiments .
The Two-Stage Nested Design
In certain multifactor experiments, the levels of
one factor (factor B) are similar but not
identical for different levels of another factor
(factor A). Such an arrangement is called a
nested or hierarchical design, with the levels of
factor B nested under the levels of factor A.

5/7/2020 1
Example
Consider a company that purchases its raw
material from three different suppliers.
The company wishes to determine whether the
purity of the raw material is the same from each
supplier. There are four batches of raw material
available from each supplier, and three
determinations of purity are to be taken from
each batch.
( The situation is shown below)

5/7/2020 2
 A two-stage nested design

5/7/2020 3
This is a two-stage nested design, with batches
nested under suppliers. this is not a factorial
experiment. If this were a factorial, then batch 1
would always refer to the same batch, batch 2 would
always refer to the same batch, and so on. This is
clearly not the case because the batches from each
supplier are unique for that particular supplier. That
is, batch 1 from supplier 1 has no connection with
batch 1 from any other supplier, batch 2 from
supplier 1 has no connection with batch 2 from any
other supplier, and so forth
5/7/2020 4
 Alternate layout for the Two-Stage
Nested design

5/7/2020 5
The linear statistical model for the two-stage nested
design is

5/7/2020 6
 there are a levels of factor A,
 b levels of factor B nested under each level of A,
and n replicates .
 The subscript j(i) indicates that the jth
level of factor B is nested under the ith level of
factor A.
 The replicates are nested within the combination
of levels of A and B( the subscript (ij )k is used for
the error term.)

5/7/2020 7
 This is a balanced nested design because
there are an equal number of levels of B
within each level of A and equal number of
replicates.
 There is no interaction between A and B,
because every level of factor B does not
appear with every level of factor A.

5/7/2020 8
• The total sum of squares can be partitioned
into a sum of squares due to factor A, a sum of
squares due to factor B under the levels of A,
and a sum of squares due to error .
SS total  SS A  SS B ( A)  SS E

5/7/2020 9
 There are (abn- 1) degrees of freedom for SST,
(a -1 )degrees of freedom for SSA,
a(b- 1) degrees of freedom for SSB(A), and
ab(n-1) degrees of freedom for error.
 Statistical software packages will perform the
analysis for a nested design. For example
Balanced ANOVA procedure in Minitab(using the
restricted model)

5/7/2020 10
 Problem-1
A rocket propellant manufacturer is studying
the burning rate of propellant from three
production processes. Four batches of
propellant are randomly selected from the
output of each process, and three
determinations of burning rate are made on
each batch. The results follow. Analyze the
data and draw conclusions

5/7/2020 11
5/7/2020 12
 Minitab output

ANOVA: Burning rate versus process, Batch

Factor Type Levels Values
process fixed 3 1 2 3
Batch(process) random 4 1 2 3 4

5/7/2020 13
Analysis of Variance for Burning rate

Source DF SS MS F P
process 2 676.06 338.03 1.46 0.281
Batch(process) 9 2077.58 230.84 12.20 0.000
Error 24 454.00 18.92
Total 35 3207.64

5/7/2020 14
Conclusion
 There is insignificant effect on burning rate
due to production process.
 Burning rate of batches in the same process
differ significantly .

5/7/2020 15
 Problem-2
The surface finish of metal parts made on four
machines is being studied. An experiment is
conducted in which each machine is run by three
different operators and two specimens from each
operator are collected and tested.
Because of the location of the machines, different
operators are used on each machine, and the
operators are chosen at random. The data are
shown in the following table. Analyze the
data and draw conclusions.

5/7/2020 16
5/7/2020 17
ANOVA: RESPO versus Machine, Operator

Factor Type Levels Values

Machine fixed 4 1 2 3 4
Operator(Machine) random 3 1 2 3

5/7/2020 18
Analysis of Variance for RESPO
Source DF SS MS F P
Machine 3 3617.67 1205.89 3.42 0.073
Operator
(machine) 8 2817.67 352.21 4.17 0.013
Error 12 1014.00 84.50
Total 23 7449.33

5/7/2020 19
Conclusion
 There is insignificant effect on surface finish
of metal parts due to machines.
 The surface finish of metal parts done by
operators in the same machine differ
significantly at 0.05.

5/7/2020 20
 Problem -3
• Consider a company that buys raw material in
batches from three different suppliers. The purity
of this raw material varies considerably, which
causes problems in manufacturing the finished
product. We wish to determine whether the
variability in purity is attributable to differences
between the suppliers. Four batches of raw
material are selected at random from each
supplier, and three determinations of purity are
made on each batch. This is, of course, a two-
stage nested design. The data are shown below.

5/7/2020 21
 Supplier -1 Supplier -2 Supplier -3

Batches 1 2 3 4 1 2 3 4 1 2 3 4

94 91 91 94 94 93 92 93 95 91 94 96

92 90 93 97 91 97 93 96 97 93 92 95

93 89 94 93 90 95 91 95 93 95 95 94

5/7/2020 22
ANOVA: RESPON versus Supplier, Batch

Factor Type Levels Values

Supplier fixed 3 1 2 3
Batch(Supplier) random 4 1 2 3 4

5/7/2020 23
 Analysis of Variance for RESPON
Source DF SS MS F P
Supplier 2 15.056 7.528 0.97 0.416
Batch(Supplier) 9 69.917 7.769 2.94 0.017
Error 24 63.333 2.639
Total 35 148.306

5/7/2020 24
Conclusion
 There is no significant effect on purity due to
suppliers .
 The purity of batches of raw material from
the same suppliers differ significantly at 0.05.

5/7/2020 25
 Split Plot Design
• In the factorial designs the combinations of
factor levels in the design could be randomly
assigned to the experimental units within a
block for a blocked factorial design, or to the
entire group of experimental units for a
completely randomized factorial design.

5/7/2020 26
• However, sometimes the levels of one or more
factors in the design are more difficult to
change or require more experimental material
to evaluate. In this situation, complete
randomization of factor-level combinations to
experimental units could make the
experiment much more time consuming or
perhaps impossible to conduct.

5/7/2020 27
• In order to ensure validity, the design of the
experiment and the analysis of the data must
take into account these different experimental
units. The designs that do this are split-plot
experiments .
The name split-plot comes from the original
application in agricultural experiments, where the
levels of some factors ( whole-plot factors) could
only be varied between plots of land , while the
levels of other factors could be varied within a
plot.

5/7/2020 28
 The linear model for the split-plot
design

5/7/2020 29
where  i ,  j , ( ) i j represent the whole plot and correspond,
respectively, to replicates(blocks),main treatments (factor A),
and whole-plot error (replicates  A) .
 k , ( ) ik , ( ) jk , ( ) ijk
represent the subplot and correspond ,respectively to the
subplot treatment (factor B),( the replicates  B),

and AB interactions, and the subplot error (replicates  AB

interaction).

5/7/2020 30
• Note that in this model
the whole-plot error is
the( replicates  A )and
the subplot error is
the three-factor interaction =(replicates 
AB interaction)
(this may not be true for other model )

5/7/2020 31
 Slightly different model for the
split-plot design

5/7/2020 32
 In this model, ( ) ij is still the whole-plot
error, but the replicates  B and
replicates  AB interactions have essentially
been pooled with εijk to form the subplot
error.
 We assume replicates are random and main
treatments and subplot treatments are fixed.

5/7/2020 33
 Problem-1
 Consider a paper manufacturer who is interested in three
different pulp preparation methods (the methods differ in the
amount of hardwood in the pulp mixture) and four different
cooking temperatures for the pulp and who wishes to study
the effect of these two factors on the tensile strength of the
paper. Each replicate of a factorial experiment requires 12
observations, and the experimenter has decided to run three
replicates. This will require a total of 36 runs. The
experimenter decides to conduct the experiment as follows .

5/7/2020 34
 A batch of pulp is produced by one of the three
methods under study. Then this batch is divided into
four samples, and each sample is cooked at one of
the four temperatures. Then a second batch of pulp
is made up using another of the three methods. This
second batch is also divided into four samples that
are tested at the four temperatures. The process is
then repeated, until all three replicates (36 runs) of
the experiment are obtained. The data are shown
below.

5/7/2020 35
5/7/2020 36
• A completely randomized factorial experiment would
require 36 batches of pulp, which is completely
unrealistic. The split-plot design requires only 9
batches total.
• In this split-plot design we have 9 whole plots, and the
preparation methods are called the whole plot or main
treatments. Each whole plot is divided into four parts
called subplots (or split-plots), and one temperature is
assigned to each. Temperature is called the subplot
treatment .

5/7/2020 37
 Solution
ANOVA: response versus ppm, temperature,
block
Factor Type Levels Values
ppm fixed 3 1 2 3
tempretu fixed 4 1 2 3 4
block random 3 1 2 3
(Balanced ANOVA,using the restricted form of the
model)

5/7/2020 38
Analysis of Variance for response

Source DF SS MS F P
ppm 2 128.389 64.194 7.08 0.049
tempretu 3 434.083 144.694 36.43 0.000
block 2 77.556 38.778 9.76 0.001
ppm*block 4 36.278 9.069 2.28 0.100
ppm*tempretu 6 75.167 12.528 3.15 0.027
Error 18 71.500 3.972
Total 35 822.972

5/7/2020 39
Conclusion
preparation methods and temperature have a
significant effect on strength, and their
interaction is significant.

5/7/2020 40
 The Split-Split-Plot Design
 The concept of split-plot designs can be
extended to situations in which randomization
restrictions may occur at any number of levels
within the experiment. If there are two levels
of randomization restrictions, the layout is
called a split-split-plot design. The following
example illustrates such a design.

5/7/2020 41
• A researcher is studying the absorption times of a
particular type of antibiotic capsule. There are
three technicians, three dosage strengths, and
four capsule wall thicknesses of interest to the
researcher. Each replicate of a factorial
experiment would require 36 observations. The
experimenter has decided on four replicates, and
it is necessary to run each replicate on a different
day.

5/7/2020 42
• Note that the days can be considered as
blocks. Within a replicate (or a block)(day), the
experiment is performed by assigning a unit of
antibiotic to a technician who conducts the
experiment on the three dosage strengths and
the four wall thicknesses.Once a particular
dosage strength is formulated, all four
wall thicknesses are tested at that strength .

5/7/2020 43
• Then another dosage strength is selected and
all four wall thicknesses are tested. Finally, the
third dosage strength and the four wall
thicknesses are tested. Meanwhile, two other
laboratory technicians also follow this plan,
each starting with a unit of antibiotic.

5/7/2020 44
• Note that there are two randomization
restrictions within each replicate (or block):
technician and dosage strength.
The whole plots correspond to the technician.
The order in which the technicians are assigned
the units of antibiotic is randomly determined.
The dosage strengths form three subplots.
Dosage strength may be randomly assigned to a
subplot .

5/7/2020 45
• Finally, within a particular dosage strength,
the four capsule wall thicknesses are tested in
random order, forming four sub-subplots. The
wall thicknesses are usually called sub-sub
treatments.
• Because there are two randomization
restrictions in the experiment the design is
called a split-split-plot design.

5/7/2020 46