Experimental Designs

1st Semester, SY 2022-2023

SLIDESMANIA.COM
 At the end of this lecture, you are expected to:
Identify the uses of RCBD and its limitations;
Layout experiments of RCBD and prepare dummy tables for
analysis and discussion;
Compute for the analysis of variance for RCBD following the
steps of hypothesis/statistical testing
SLIDESMANIA.COM
 At the end of this lecture, you are expected to:
Identify the uses of latin square design and its limitations;
Layout experiments of latin square design and prepare
dummy tables for analysis and discussion; and
Compute for the analysis of variance for latin square design
following the steps of hypothesis/statistical testing
SLIDESMANIA.COM
 The use of one-way classification or CRD requires that the eu's are
homogeneous. Sometimes this requirement is not satisfied since the
eu's are markedly heterogeneous with respect to some criteria of
classification. For instance, plots differ in fertility, trees differ in
age or height, analysts differ in efficiency, etc.
SLIDESMANIA.COM
 The differences among the eu's is a major source of experimental
error. When the eu’s are heterogeneous, the appropriate design must
account for these heterogeneities. These designs are the

randomized complete block design (RCBD) and the latin

square design.
SLIDESMANIA.COM
 THE RANDOMIZED
COMPLETE BLOCK DESIGN
(RCBD)
SLIDESMANIA.COM
 ▪ As a local control, one way of increasing the precision of an
experiment is by proper grouping or blocking of the
experimental units.
▪ The eu's are grouped into r blocks in such a way that the
differences between the units among different blocks are greater
than the differences between the units within each block.
SLIDESMANIA.COM
 ▪ Likewise, the blocking should be done in such a way that the blocks
cut across are perpendicular to the direction of the eu's gradient.
▪ This way, if there are differences among the blocks, the
variability is removed from the experimental error thereby
improving the precision of the experiment.
SLIDESMANIA.COM
 Description of the Design
✓ Probably the most used and useful of the experimental designs.
✓ Takes advantage of grouping similar experimental units into
blocks or replicates.
✓ The blocks of experimental units should be as uniform as
possible.
SLIDESMANIA.COM
 Description of the Design
✓ The purpose of grouping experimental units is to have the units
in a block as uniform as possible so that the observed
differences between treatments will be largely due to “true”
differences between treatments.
SLIDESMANIA.COM
 Advantages of RCBD
✓ Generally, more precise than the CRD.
✓ No restriction on the number of treatments or replicates.
✓ Some treatments may be replicated more times than others.
✓ Missing plots are easily estimated.
SLIDESMANIA.COM
 Advantages of RCBD
✓ Whole treatments or entire replicates may be deleted from the
analysis.
✓ If experimental error is heterogeneous, valid comparisons can
still be made.
SLIDESMANIA.COM
 Disadvantages of RCBD
✓ Error df is smaller than that for the CRD (problem with a small
number of treatments).
✓ If there is a large variation between experimental units within a
block, a large error term may result (this may be due to too
many treatments).
SLIDESMANIA.COM
 Disadvantages of RCBD
✓ If there are missing data, a RCBD experiment may be less
efficient than a CRD

NOTE: The most important item to consider when choosing a design is the
uniformity of the experimental units.
SLIDESMANIA.COM
 Randomization and Layout
SLIDESMANIA.COM
 Randomization and Layout
SLIDESMANIA.COM
 Randomization and Layout
SLIDESMANIA.COM
 Data Presentation
SLIDESMANIA.COM
 Linear Model
𝑌𝑖𝑗 = 𝜇 + 𝜏𝑖 + 𝜌𝑗 + 𝜀𝑖𝑗 𝑖 = 1, 2, … , 𝑡; 𝑗 = 1, 2, … , 𝑟
where: 𝜇 = general mean of all possible observation;
𝜏𝑖 = effect of the 𝑖𝑡ℎ treatment;
𝜌𝑗 = 𝑗𝑡ℎ block effect;
𝜀𝑖𝑗 = random error associated with the 𝑗𝑡ℎ block of the 𝑖𝑡ℎ
treatment.
SLIDESMANIA.COM

𝜇𝑖 = μ + 𝜏𝑖 , the true mean of the 𝑖𝑡ℎ treatment

 Analysis of Results
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Statistical hypothesis
Ho: All variety means are not different
Ha: At least two variety means are different.
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Test statistics and critical value
𝑀𝑆𝑇𝑟
Test statistics: 𝐹𝑐 = 𝑀𝑆𝐸
;
Critical Value: 𝐹𝑡 = 𝐹𝛼 𝑡−1 , 𝑟−1 𝑡−1

= 𝐹0.05 3,12
= 3.49
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Decision rule: Reject Ho (and accept Ha) if 𝐹𝑐 ≥ 𝐹𝑡 ; else, accept Ho.
✓ Construct the ANOVA table outlined as follows:
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Computation: notations
Let 𝑌𝑖𝑗 = observation on the jth block with the ith treatment
𝑌𝑖. = total observation with the ith treatment
𝑌.𝑗 = total observation on the jth block
𝑌.. = grand total of all observations
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the sum of squares
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the sum of squares
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the mean squares
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the test statistics
𝑀𝑆𝑇𝑟 54.800
𝐹𝑐 = = = 36.36
𝑀𝑆𝐸 1.507
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ Set up the ANOVA table
SLIDESMANIA.COM
 Analysis of Results
1. Testing significance of treatments via ANOVA F-test
✓ State the decision and make a conclusion.
Decision: Since 𝐹𝑐 ≥ 𝐹𝑡 , we reject the Ho and accept the Ha.
Conclusion: At 5% level of significance, there are at least two
varieties which gave significantly different yield.
SLIDESMANIA.COM
 Analysis of Results
2. Other summary statistics
✓ Coefficient of variation, CV(%)

𝑀𝑆𝐸 1.507
𝐶𝑉 % = × 100% = × 100% = 3.76%
ത .
𝑌. 32.66

𝑌.. 653.2
Note that 𝑌. . = 𝑟×𝑡 = 5×4 = 32.66
ത
SLIDESMANIA.COM
 Analysis of Results
2. Other summary statistics
✓ Standard error of a treatment mean, 𝑠. 𝑒. 𝑌
ത𝑖.

𝑀𝑆𝐸 1.507
𝑠. 𝑒. 𝑌ത𝑖. = = = 0.55
𝑟 5
SLIDESMANIA.COM
 Analysis of Results
2. Other summary statistics
✓ Standard error of the difference between two treatment means,
𝑠. 𝑒. 𝑑ҧ

2𝑀𝑆𝐸 2 1.507
𝑠. 𝑒. 𝑑ҧ = = = 0.78
𝑟 5
SLIDESMANIA.COM
 Analysis of Results
3. Multiple comparisons among means using DMRT. At 𝛼 = 0.05,
SLIDESMANIA.COM
 Analysis of Results
3. Multiple comparisons among means using DMRT. At 𝛼 = 0.05,
SLIDESMANIA.COM
 Relative efficiency of RCBD over CRD
𝑟 − 1 𝑀𝑆𝑅 + 𝑟 𝑡 − 1 𝑀𝑆𝐸
R.E. (RCBD/CRD) = × 100%
𝑟𝑡 − 1 𝑀𝑆𝐸
5 − 1 6.725 + 5 4 − 1 1.507
= × 100%
20 − 1 1.507
= 172.89%
SLIDESMANIA.COM
 Relative efficiency of RCBD over CRD
▪ There is appreciable gain in efficiency in using RCBD over CRD; that
is, RCBD is 72.89% more efficient than CRD of the same number of
replications.
▪ To be of equal efficiency as that of an RCBD of 5 replications, the
CRD requires (1.7289)(r) = (1.7289)(5) = 9 replications
SLIDESMANIA.COM
 LATIN SQUARE DESIGN
SLIDESMANIA.COM
 Facts About the LS Design
✓ The LS design, also a complete block design, is a balanced two-
way classification design. This is used when the experimental
units can be grouped into two categories (double blocking) such
that the variations due to each category are eliminated from the
experimental error.
SLIDESMANIA.COM
 Facts About the LS Design
✓ One grouping or blocking is denoted as the column
classification and the other as the row classification.
The grouping of the eu's should be done so that the differences
among rows and columns represent major sources of variation.
SLIDESMANIA.COM
 Facts About the LS Design
✓ The LS design may be used to study three factors
simultaneously assuming that the factors have no interactions
among themselves.
✓ Each row contains every treatment.
✓ Each column contains every treatment.
SLIDESMANIA.COM
 Advantages of the LS Design
✓ You can control variation in two directions.
✓ It is possible to increase efficiency as compared to the RCBD.
SLIDESMANIA.COM
 Disadvantages of the LS Design
✓ The number of treatments must equal the number of replicates.
✓ The experimental error is likely to increase with the size of the
square.
✓ Small squares have very few degrees of freedom for
experimental error.
SLIDESMANIA.COM
 Disadvantages of the LS Design
✓ You can’t evaluate interactions between:
a. Rows and columns
b. Rows and treatments
c. Columns and treatments.
SLIDESMANIA.COM
 Blocking, Randomization and Layout
▪ In the LS design, the whole experimental materials are divided
into as many rows and as many columns as there are many
treatments.
▪ The treatments are then arranged in blocks in two ways: by rows
and by columns such that each treatment occurs once in each row
and column.
▪ Each row is a complete block, each column is likewise a complete
block.
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Randomization Procedure:
1. Obtain a basic 𝑡 × 𝑡 latin square plan.
2. Randomize the assignment of the row classifications to the
rows on the plan.
3. Randomize the assignment of the column classifications to the
columns on the plan.
4. Randomize the assignment of treatments to the treatments on
the plan.
SLIDESMANIA.COM

5. Rearrange the rows and columns in sequence

 Blocking, Randomization and Layout
Illustration:
An experiment in a 4 × 4 latin square design is to be
conducted to measure the pesticide residue levels in dried fish of
different kinds (𝑇1 , 𝑇2 , 𝑇3 , 𝑇4 ). It is suspected that the source
locations (𝑆1 , 𝑆2 , 𝑆3 , 𝑆4 ) and the chemists (𝐶1 , 𝐶2 , 𝐶3 , 𝐶4 ) that
perform the chemical analyses contribute to the variation of the
results and these sources of variation have to be accounted. It is
assumed that the source locations, the chemist, and kinds of dried fish
have no interactions.
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:

Treatments: 𝑇1 , 𝑇2 , 𝑇3 , 𝑇4
Row classification: 𝑆1 , 𝑆2 , 𝑆3 , 𝑆4
Column classification: 𝐶1 , 𝐶2 , 𝐶3 , 𝐶4
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:
Step 1: Obtain Basic 4 × 4 latin square plan
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:
Step 2: Randomize the assignment of the source location to the rows
Draw sequence: 1 2 3 4 = source locations
Drawn numbers : 2 4 1 3 = plan rows
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:
Step 3: Randomize the assignment of the chemists to the columns
Draw sequence: 1 2 3 4 = chemists
Drawn numbers : 4 2 1 3 = plan columns
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:
Step 4: Randomize the assignment of the treatments
Draw sequence: 1 2 3 4 = dried fish
Drawn letters : B D C A = plan letters
SLIDESMANIA.COM
 Blocking, Randomization and Layout
Illustration:
Step 5: Rearrange the rows and columns in sequence
SLIDESMANIA.COM
 Data Presentation
Let Yij(k) = observation on the response variable from the experimental
unit on the ith row and jth column with the kth treatment
SLIDESMANIA.COM
 Linear Model
𝑌𝑖𝑗(𝑘) = 𝜇 + 𝜌𝑖 + 𝛾𝑗 + 𝜏𝑘 + 𝜀𝑖𝑗(𝑘)
where: 𝜇 = general mean of all possible observation;
𝜌𝑖 = the ith row effect; 𝛾𝑗 = the jth column effect;
𝜏𝑘 = the effect of the kth treatment;
𝜀𝑖𝑗 = random error associated with the ijth eu with the kth
treatment;
𝜇𝑖 = μ + 𝜏𝑖 , the true mean of the kth treatment
SLIDESMANIA.COM
 Analysis of Results
1. Testing of significance
The preliminary test of significance of the treatment means, row
means, column means is done via ANOVA
The outline of the ANOVA table is
SLIDESMANIA.COM
 Analysis of Results
1. Testing of significance
a. Test the significance of the treatments
Test Statistics : Fc = MSTr/MSE vs Ft = Fα[(t-1), (t-1)(t-2)]
Decision Rule : Reject Ho if Fc ≥ Ft ⇒ at least two treatment means
are different
SLIDESMANIA.COM
 Analysis of Results
1. Testing of significance
b. Test the significance of the row classifications
Test Statistics : Fc = MSR/MSE vs Ft = Fα[(t-1), (t-1)(t-2)]
Decision Rule : Reject Ho if Fc ≥ Ft ⇒ at least two row means are
different
SLIDESMANIA.COM
 Analysis of Results
1. Testing of significance
c. Test the significance of the column classifications
Test Statistics : Fc = MSC/MSE vs Ft = Fα[(t-1), (t-1)(t-2)]
Decision Rule : Reject Ho if Fc ≥ Ft ⇒ at least two column means are
different
SLIDESMANIA.COM
 Illustration
An experiment in a latin square design was conducted to compare the potassium
contents of four varieties of a composting grass. The data are summarized below:
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Critical value
Critical Value: 𝐹𝑡 = 𝐹𝛼 𝑡−1 , 𝑡−1 𝑡−2

= 𝐹0.05 3,6
= 4.76
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the sum of squares
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the sum of squares
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Computation: compute the mean sum of squares
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Set-up ANOVA Table
SLIDESMANIA.COM
 Illustration
1. Testing significance of treatments via ANOVA F-test
✓ Conclusion:
The ANOVA indicated that there are significant differences
among the means of the different sources. It is also indicated
that there are significant differences among the varieties.
However, the ANOVA fail to provide sufficient evidence that there
are significant differences among the means of the chemists.
SLIDESMANIA.COM
 Illustration
2. Other summary statistics
✓ Coefficient of variation, CV(%)

𝑀𝑆𝐸 0.051
𝐶𝑉 % = × 100% = × 100% = 36.45%
ത
𝑌. . 0.6196

𝑌.. 9.914
Note that 𝑌. . = 2 = 2 = 0.6196
ത
𝑡 4
SLIDESMANIA.COM
 Illustration
2. Other summary statistics
✓ Standard error of a treatment mean, 𝑠. 𝑒. 𝑌
ത..(𝑘)

𝑀𝑆𝐸 0.051
𝑠. 𝑒. 𝑌ത𝑖. = = = 0.113
𝑟 4

Note that in LS design, 𝑟 = 𝑡. Thus, 𝑟 = 4.

SLIDESMANIA.COM
 Assignment 6.1
Perform DMRT at 𝛼 = 5% for pairwise mean comparison among
a. Varieties
b. Source Locations
SLIDESMANIA.COM
 Credits
Presentation Template: SlidesMania
Please keep this slide or mention us in the footer of a slide.
SLIDESMANIA.COM