COMPLETELY RANDOMIZED DESIGN

Designing an experiment means deciding how the observations or measurements should be taken
to answer a particular question in a valid, efficient and economical way. A well designed
experiment helps the workers to properly partition the variation of the data into respective
component in order to draw valid conclusion.
The Completely Randomized Design (CRD) is the simplest of all the design based on
randomization and replication. In CRD, all treatments are randomly allocated among all
experimental subjects. It relies on the randomization to control for the effect of extraneous
variables other than treatments. CRD is for the studying the effect on the primary without the
need to take other nuisance variables into account.
This allows every experimental unit; i.e.; plot, animal, soil sample, genotype etc., to have an
equal probability of receiving a treatment.
Advantages of CRD:
1. It is simplest of the all experimental designs.
2. Easy to layout and design.
3. There is complete flexibility in the number of treatments and number of replications. This
means that any number of treatments and any number of units per treatment may be used.
4. The number of replications need not to be same for each treatment.
5. The CRD provides maximum degree of freedom (df) of experimental error for the
experiment.
Disadvantages of CRD:
1. Principle of the local control is not employed in this experimental design.
2. Experimental error is inflated as it encompasses entire variation present among
experimental units.
3. Results of CRD are reliable only when experimental units are nearly homogeneous
When to use CRD?
1. CRD is used when experimental units are almost homogeneous, that is the variation
among experimental units is small e.g. under green house or laboratory conditions.
2. In small experiments where there is small degree of freedom.
3. When experimental units are the part of the thoroughly mixed chemical or powder or
when a quantity of material is thoroughly mixed and then divided into small lots to from
the experimental units to which treatments are randomly assigned.
Example of CRD:
Federal Seed Certification and Registration Department (FSC&RD) issues seed certification tags
of different categories (Pre-basic, basic, and certified) to the public and private sector seed
producing entities. Before issuance of the certification tag, FSC&RD checks various seed quality
parameters on of which is seed germination percentage.
Suppose, you are incharge of the seed quality lab of FSC&RD. You received wheat seed samples
of variety “Akbar 19” from 10 seed companies (Company A-J) to test germination percentage.
You have the facility of a controlled germination room to test germination percentage in petri
dishes. How would you proceed?
Step 1: Decide your experimental design
You will choose ……………………………………… design based upon the following reasons.
1. ………………………………………………………………………….
2. ………………………………………………………………………….
3. ………………………………………………………………………….
Step 2: Identify treatments and their number
In this particular example, source of seed i.e. seed companies are the treatments and total number
of treatments (t) is 10.
Number of treatments ( t )=10
Step 3: Decide number of replications
You decided to repeat each treatment three times, hence, number of replications (r) is three.
Number of replications ( r )=3
Step 4: Identify experimental units and calculate their number
In this particular case, one petri dish is equivalent to one experimental unit. Total number of
experimental units (n) required is calculated by multiplying number of treatments (t) with
number of replications (r).

Number of experimental units=Number of treatments × Number of replications

n=t ×r
n=10 ×3=30

Hence, you need 30 petri dishes to perform this experiment.

Step 5: Make layout of your experiment
Layout plan of your experiment depends upon the experimental design you choose in step 1. In
this particular case, you choose Completely Randomized Design to perform the experiment and
you will make layout plan of your experiment accordingly. In this case, treatments will be
allotted to 30 petri dishes (experimental units) randomly. For purpose of randomization, you may
use lottery method, random number table method or random number generator function of
calculator. Suppose you choose to use lottery method in this particular instance.
i. Write down all of treatments with replication number on pieces of papers / slips

Company Company Company Company Company Company

A A A B B B
R1 R2 R3 R1 R2 R3

Company Company Company Company Company Company

C C C D D D
R1 R2 R3 R1 R2 R3

Company Company Company Company Company Company

E E E F F F
R1 R2 R3 R1 R2 R3

Company Company Company Company Company Company

G G G H H H
R1 R2 R3 R1 R2 R3

Company Company Company Company Company Company

I I I J J J
R1 R2 R3 R1 R2 R3

ii. Fold the slips so that you cannot see the name of treatments written on them and mix them in
a bowl.
iii.Take thirty (30) petri dishes and label them from one to thirty.

1 2 3 4 5

6 7 8 9 10

11 12 13 14 15

16 17 18 19 20

21 22 23 24 25

26 27 28 29 30

iv. Pick slips from the bowl one by one and assign treatments to the petri dishes accordingly.
Suppose the 1st slip you picked from the bowl has “Company C R1” written on it, so petri
dish number 1 will have 1 st replication of seed received from company C. The 2 nd slip you
picked has “Company A R2 written on it, so the 2 nd Petri dish will contain 2nd replication of
the seed received from company A. Repeat this process for all thirty slips so that each
experimental unit (petri dish) has equal chance of receiving any treatment and all thirty
treatments are assigned randomly to all experimental units.
 1 C-R1 2 A-R2 3 H-R1 4 D-R2 5 E-R2

6 D-R1 7 G-R3 8 B-R2 9 A-R3 10 G-R2

11 A-R1 12 E-R1 13 C-R2 14 I-R3 15 F-R3

16 F-R1 17 B-R1 18 G-R1 19 J-R2 20 J-R1

21 H-R2 22 C-R3 23 I-R1 24 F-R2 25 J-R3

26 E-R3 27 D-R3 28 I-R2 29 B-R3 30 H-R3

Step 6: Experiment and data recording

Perform the experiment and record the data for germination percentage of wheat seed samples
received from different seed companies on your data book.

Petri Sample G (%) Petri Sample G (%) Petri Sample G (%)

Dis # Dis # Dis #
1 C-R1 91 11 A-R1 87 21 H-R2 93
2 A-R2 87 12 E-R1 51 22 C-R3 92
3 H-R1 95 13 C-R2 94 23 I-R1 96
4 D-R2 89 14 I-R3 98 24 F-R2 52
5 E-R2 46 15 F-R3 50 25 J-R3 63
6 D-R1 89 16 F-R1 56 26 E-R3 49
7 G-R3 61 17 B-R1 90 27 D-R3 87
8 B-R2 94 18 G-R1 63 28 I-R2 98
 9 A-R3 86 19 J-R2 69 29 B-R3 91
10 G-R2 65 20 J-R1 65 30 H-R3 93

Step 7: Data Analysis according to CRD Design

i. Define your hypothesis

Null hypothesis: Ho: Germination of seed samples received from all of 10 seed
companies is same
There is no difference in germination percentage of seed
samples received from 10 seed companies
All seed samples have same germination percentage
All means are the same

Alternate hypothesis: HA: At least two seed samples have germination percentage
different from each other
All means are not same

ii. Arrange the data into treatment × replication matrix. Calculate treatments total and treatments
mean for further data analysis.
Treatment Total (∑ Y i)=R 1+ R 2+ R 3
And

Treatment Mean (Y̅ i)=∑ Y i/r

Whereas;
Yi is YA, YB, YC, …..YJ

Y̅ I is Y̅ A, Y̅ B, Y̅ C, ….. Y̅ J

e.g. Treatment total for Company A will be

∑YA = 87 + 87 + 68 = 260
And treatment mean for Company A will be

Y̅ A = 260/3 = 86.7

A B C D E F G H I J
R1 87 90 91 89 51 56 63 95 96 65
R2 87 94 94 89 46 52 65 93 98 69
R3 86 91 92 87 49 50 61 93 98 63
Total 260 275 277 265 146 158 189 281 292 197
 Mean 86.7 91.7 92.3 88.3 48.7 52.7 63.0 93.7 97.3 65.7

The Grand Total for all treatments will be

Grand Total = GT = ∑YA + ∑YB+ ∑Yc+ ………. + ∑YJ

GT = 260+275+277+ ………. +197

GT = 2340

Likewise, The Grand Mean will be

Grand Mean = GM = GT/n

GM = 2340 / 30

GT = 78

iii.Calculate Correction Factor (CF)

Correction Factor = CF = (GT)2/n

CF = (2340)2 / 30

CF = 182520

iv. Calculate Total Sum of Square (SSTotal)

Total Sum of Sqaure = SSTotal = ∑Yij2 - CF

SSTotal = (87)2 + (90)2 + (91)2 + 90 + ….. + (63)2 - 182520

SSTotal = 191818 – 182520 = 9298

v. Calculate Treatment Sum of Square (SSTreatment)

2
∑ yi
Treatment Sum of Sqaure = SSTreatment = −CF
r
2 2 2 2
(260) (275) (277) (197)
SSTreatment = + + + …….. + -182520
3 3 3 3

SSTreatment = 191738 – 182520 = 9218

vi. Calculate Error Sum of Square (SSError)
Error Sum of Sqaure = SSError = SSTotal - SSTreatment

SSError = 9298 – 9218 = 80

vii. Make ANOVA Table

Source of df SS MS FCalculated FTabulated
Variation
Treatments t-1 = 10-1 = 9 9218 SS 9218 1024.2 2.39(5%)
= =
df 9 4 3.46(1%)
=1024.2 256.05
Error (n-1)-(t-1) 80 SS 80
= =4
=29-9=20 df 20
Total n-1 = 30-1 =29 9298

viii. Draw inference from ANOVA table

If;
FCalculated < FTabulated Alternate hypothesis is rejected

If;
FCalculated > FTabulated Alternate hypothesis is not rejected

In our example
FCalculated > FTabulated
256.05>1.96 It means that germination percentage of at least two seed
samples is significantly different from each other.

It is important to note that significant F test only tells that at least one of the treatments is
different from others.

ix.Calculate Coefficient of variability (CV)

¿ √ EMS ×100
Coefficient of Variability = CV Grand Mean

¿ √ ×100
4
78
¿
√ 4 ×100=2.56
78

x. Calculate Standard Errors (SE) of treatment means

 Standard Error of a particular treatment mean = SE(Y̅ i) ¿
√ EMS
ri

SE(Y̅ A)= SE(Y̅ B)= SE(Y̅ C)= ………………...…..= SE(Y̅ J) ¿

√ 4
3
=1.15

xi.Calculate Standard Error (SE) of difference between two treatment means

SE (Y̅ A - Y̅ B)
√
¿ EMS
( r1 + r1 )
A B

√(
¿ 4
1 1
+
3 3 ) √
¿ 4 ( 23 ) ¿ √ 2.67 = 1.63

Step 8: Post Hoc Test

i. Tuckey’s HSD (Honestly Significant Difference) Test
a. Calculate HSD value

HSD value
=
Q √
¿ q α (t , error df )
EMS 1 1
+
(
2 ri r j )
¿ 5.01
√ ( )
4 1 1
+
2 3 3
¿ 5.01
√()
4 2
2 3
¿ 5.01 ×1.15 = 5.76

b.Arrange treatments in ascending order of means

I H C B D A J G F E
97.3 93.7 92.3 91.7 88.3 86.7 65.7 63.0 52.7 48.7
c. Make comparison table
Difference between two treatments is considered significant if difference between
treatments means is greater than HSD value.
Comparison Means Difference HSD Value Result
I vs H 97.3-93.7 = 3.6 5.76 Non-Significant
I vs C 97.3-92.3 = 5.0 5.76 Non-Significant
I vs B 97.3-91.7 = 5.6 5.76 Non-Significant
I vs D 97.3-88.3 = 9.0 5.76 Significant
I vs A 97.3-86.7 = 10.6 5.76 Significant
I vs J 97.3-65.7 = 31.6 5.76 Significant
I vs G 97.3-63.0 = 34.3 5.76 Significant
I vs F 97.3-52.7 = 44.6 5.76 Significant
I vs E 97.3-48.7 = 48.6 5.76 Significant
H vs C 93.7-92.3 = 1.4 5.76 Non-Significant
H vs B 93.7-91.7 = 2.0 5.76 Non-Significant
H vs D 93.7-88.3 = 5.4 5.76 Non-Significant
H vs A 93.7-86.7 = 7.0 5.76 Significant
H vs J 93.7-65.7 = 28.0 5.76 Significant
H vs G 93.7-63.0 = 30.7 5.76 Significant
H vs F 93.7-52.7 = 41.0 5.76 Significant
H vs E 93.7-48.7 = 45.0 5.76 Significant
C vs B 92.3-91.7 = 0.6 5.76 Non-Significant
C vs D 92.3-88.3 = 4.0 5.76 Non-Significant
C vs A 92.3-86.7 = 5.6 5.76 Non-Significant
C vs J 92.3-65.7 = 26.6 5.76 Significant
C vs G 92.3-63.0 = 29.3 5.76 Significant
C vs F 92.3-52.7 = 39.6 5.76 Significant
C vs E 92.3-48.7 = 43.6 5.76 Significant
B vs D 91.7-88.3 = 3.4 5.76 Non-Significant
B vs A 91.7-86.7 = 5.0 5.76 Non-Significant
B vs J 91.7-65.7 = 26.0 5.76 Significant
B vs G 91.7-63.0 = 28.7 5.76 Significant
B vs F 91.7-52.7 = 39.0 5.76 Significant
B vs E 91.7-48.7 = 43.0 5.76 Significant
D vs A 88.3-86.7 = 1.6 5.76 Non-Significant
D vs J 88.3-65.7 = 22.6 5.76 Significant
D vs G 88.3-63.0 = 25.3 5.76 Significant
D vs F 88.3-52.7 = 35.6 5.76 Significant
D vs E 88.3-48.7 = 39.6 5.76 Significant
A vs J 86.7-65.7 = 21.0 5.76 Significant
A vs G 86.7-63.0 = 23.7 5.76 Significant
A vs F 86.7-52.7 = 34.0 5.76 Significant
A vs E 86.7-48.7 = 38.0 5.76 Significant
J vs G 65.7-63.0 = 2.7 5.76 Non-Significant
J vs F 65.7-52.7 = 13.0 5.76 Significant
J vs E 65.7-48.7 = 17.0 5.76 Significant
G vs F 63.0-52.7 = 10.3 5.76 Significant
G vs E 63.0-48.7 = 14.3 5.76 Significant
F vs E 52.7-48.7 = 4.0 5.76 Non-Significant