Lecture Note 2

STA 433
Statistical Methods &
Field Experiments
Dr. M. K. Garba Caution
This learning material is NOT FOR SALE by
Department of Statistics, anybody, not even by the Class
University of Ilorin, Nigeria Representative. Appropriate sanctions
08035568461 await the violators.
garba.mk@unilorin.edu.ng
 DESIGN OF
EXPERIMENTS
&
ANALYSIS OF
VARIANCE (ANOVA)
This learning material is NOT FOR SALE 2
 Analysis Of Variance (ANOVA)
The analysis of variance is a procedure used to
investigate whether differences exist among
more than two population means.
This technique is an F-test and uses the
F-statistical tables with v1 and v2 degrees of
freedom (d.f. for numerator and denominator
respectively) at a specified level of significance.

This learning material is NOT FOR SALE 3

 Completely Randomized Design
This is the simplest design in which it is believed
that the only source of variability in the
experimental units is the random (non-assignable)
variation. Thus, there is no need for error control
at the design stage of the experiment.
To use this design, it is necessary that all
experimental units are as similar as possible
(homogenous) in all aspects.
This learning material is NOT FOR SALE 4
In CRD, allocation of treatments to experimental
units is completely random. The randomization can
be done using method of random numbers or hat
method.
Advantages and Disadvantage of CRD
CRD is flexible to design and implement. It has
higher degree of freedom for error which improves
precision.
The major disadvantage of this design is that it
becomes inefficient if the units are not homogenous.
This learning material is NOT FOR SALE 5
In CRD, apart from the random variation, the
only assignable cause of variation is that due to
treatments. Therefore, the total variation in this
experiment can be expressed as

Total variation = Variation due to treatments

+ Random variation

The Completely Randomized Design (CRD) is

also knows as One-Way ANOVA in which only
one factor is considered in the experiment.
This learning material is NOT FOR SALE 6
 Layout of Responses from CRD
Treatment
Replication 1 2 3 . . . t
1 y11 y21 y31 . . . yt1
2 y12 y22 y32 . . . yt2

Grand Total =
3 y13 y23 y23 . . . yt3
. . . . .
. . . . .
. . . . .
r y1r y2r y3r . . . ytr
Totals y1. y2. y3. . . . yt. y. .
The design consists of t treatments replicated r times 7
This learning material is NOT FOR SALE
 Model for CRD
The model for CRD ANOVA is given as

yij = μ + ti + eij
where
yij is the observed response for treatment i in rep j
μ is the overall mean
ti is the effect of treatment i
eij is the random error term for treatment i in rep j
8
This learning material is NOT FOR SALE
 ANOVA TABLE
To examine the differences among effects of the
treatments, an ANOVA Table as presented below
is utilized
Source of Degrees Sum of Mean Square F-cal
Variation of Squares
freedom
Treatment t–1 SSt MSt = SSt/t-1 F = MSt/MSE

Error n–t SSE MSE = SSE/n-t

Total n–1 SST

9
 Computations of Sums of Squares

𝟐 𝟐 𝟐
𝒚𝒊. 𝒚. . 𝟐 𝒚. .
𝑺𝑺𝒕 = − and 𝑺𝑺𝑻 = 𝒚𝒊𝒋 −
𝒓 𝒏 𝒏

where SSE = SST – SSt 𝒂𝒏𝒅 𝒚. . = 𝒚𝒊𝒋

10
 Statement of Hypotheses

H0: μ1 = μ2 = μ3 = . . . μt
(That is, all treatment effects are the same)

versus

H1: μ1 ≠ μ2 ≠ μ3 ≠ . . . μt
(i.e. at least one of the effects differs from others)

This learning material is NOT FOR SALE 11

 Decision Rule
Reject the null hypothesis (H0) if the Fcalculated is
greater than Ftabulated (Fv1, v2, α).
where v1 is the degrees of freedom for treatment
and v2 is the degrees of freedom for error

This learning material is NOT FOR SALE 12

 Example 1
A researcher wished to try different techniques to lower the
cholesterol level of cattle reared in the barn managed by the
Department of Animal Production. The subjects were randomly
assigned to three groups each with a technique from medication,
exercises and diets. After four weeks, the reduction in cholesterol
are recorded as shown below. At α = 0.05, test the claim that there
is no difference among the means of the techniques.
Medication Exercise Diet
10 6 5
12 8 9
9 3 12
15 0 8
13 2 4
13
This learning material is NOT FOR SALE
 Example 2
An experiment was carried out to determine if the yields of five
different varieties of cowpea tagged A, B, C, D & E differ. Each
variety are replicated on four plots of land and the yields in tons
are given below. Assume that the plots of land in each location are
of similar fertility and the varieties were assigned at random to the
plots, determine if there is significant difference in the yields at 5%
significance level.
Varieties Yields
A 20 12 15 19
B 17 14 12 15
C 23 16 18 14
D 15 17 20 12
E 21 14 17 18
14
This learning material is NOT FOR SALE
 Example 3
Copy and complete the following abridged ANOVA
table. State the relevant hypotheses and carry out the
appropriate statistical test.
Source df SS MS F

Treatment 4

Error 0.1882

Total 39 16.8528

15
This learning material is NOT FOR SALE
 UNEQUAL REPLICATIONS
In reality, situation may arise where equal replications
for the treatments under consideration are not feasible or
such experiment is carried out purposefully such that
experimental materials are not equally replicated for the
treatments. The analysis of data from such experiment is
done similar to the case of equal replications learnt early
except for the computation of sum of squares for
treatment (SSt) where the number of replications for
each treatment (ri) is used to divide square of sum of
responses for each treatment instead of just r.
16
This learning material is NOT FOR SALE
 Example 4
The scores in STA433 for four different sessions
were randomly selected and recorded below. Test at
5% significance level if significant differences
exist in the scores.
2016/2017 2017/2018 2018/2019 2019/2020
72 81 88 74
80 74 82 71
83 77 90 77
75 87 70
80

17
This learning material is NOT FOR SALE
 Attempt these questions
#1
In an experiment to investigate the performance of four
different types of growth hormones in plants. Each hormone
was applied to six different plants of the same species and the
growth was observed after three weeks of application and
observations recorded. Complete the resulting ANOVA table
and carry out F-test at 0.05 level of significance.

Source of variation df Sum of Squares Mean Square F

Hormone
Error 6.5083
Total 512.9583

18
This learning material is NOT FOR SALE
#2 An experiment involves 4 pens consisting of 10 rats each.

A pen was given a diet with different carbohydrate for a

period of 30days. At the end of the experiment, the DNA
content of the liver of each rat was determined (mg/g
liver) and recorded. The following summaries were
Obtained from the recorded data.
y1. = 260.8, y2. = 246.9, y3. = 299.5, y4. = 338.4,
y.. = 1145.6 and ∑∑yij2 = 33882.24
At 5% level of significance, carry out a test to verify if
the data indicate that the average DNA content is affected
by the type of carbohydrate in the diet?
19
This learning material is NOT FOR SALE
#3
The following dataset was obtained from an experiment to test the
effects of four feeds (W, X, Y & Z) on increase in weight of
animals. Six animals were randomly fed with each of the four feeds:
W X Y Z
5.2 6.5 5.8 8.3
4.5 8.0 4.7 6.1
6.0 6.1 6.4 7.8
6.1 7.5 4.9 7.0
6.7 5.9 6.0 5.5
5.8 5.6 5.2 7.2
Test whether there is significance difference in the effects of
the feeds at 5% significance level.

20
This learning material is NOT FOR SALE
 Randomized Complete Block
Design (RCBD)
In many experimental problems, it is necessary
to design the experiment such that variability
arising from extraneous sources can be
systematically controlled.
RCBD may be used when the experimental units
can be grouped in a way that the number of units
in a group is equal to the number of treatments.
This learning material is NOT FOR SALE 21
The groups are referred to as blocks and they are used
to control or eliminate a known source of variability
which in turn improves the accuracy of the experiment
by reducing the experimental error.
The purpose of grouping or blocking is to have units in
a group as homogenous as possible so that observed
differences in a group are mainly due to treatments.
In RCBD, the number of plots per block equals the
number of treatments, and the blocks are of equal size.
This is why the design is known as Complete Block
Design.
This learning material is NOT FOR SALE 22
Advantages and Disadvantages of RCBD
1. Where there is one source of variability in the units, the RCBD
gives a higher precision that the CRD.
2. The layout of the design is simple & flexible as there is no
restriction on the number of treatments & blocks except
limitation in terms of resources for carrying out the experiment
or in getting enough similar units for each block.
3. When there is a missing value, the statistical analysis for
estimating it is simple.
The prominent disadvantages are highlighted as follows
1. RCBD requires equal number of replications for each
treatment. It may not be an easy task to secure enough similar
units to form complete block when the number of treatments is
large.
2. The design is not effective in controlling experimental error if
there are two sources of variability or gradients.
23
This learning material is NOT FOR SALE
Layout of Responses from RCBD
Blocks
Treatments 1 2 3 . . . b Total
1 y11 y21 y31 . . . yb1 T1
2 y12 y22 y32 . . . yb2 T2
3 y13 y23 y23 . . . yb3 T3
. . . . . .
. . . . . .
. . . . . .

t y1t y2t y3t . . . ybt Tt

Total B1 B2 B3 . . . Bb GT
24
This learning material is NOT FOR SALE
In RCBD, the assignable causes of variation are
those due to the blocks and treatments. Hence,
the total variation can be expressed as
Total variation = Block variation + Treatment
variation + Random variation
Accordingly, the observed value from a unit can
be expressed as
Observed value = Overall mean + Block effect
+ Treatment effect + Random error

This learning material is NOT FOR SALE 25

 Model for RCBD
The model for RCBD ANOVA is
yij = μ + bi + tj + eij
For i = 1, 2, 3, . . ., t and j = 1, 2, 3, . . . , b, where
yij is the observed response for the unit in block i
receiving treatment j
μ is the overall mean
bi is the effect of block i
tj is the effect of treatment j
eij is the random error term for the unit in block i
receiving treatment j
26
This learning material is NOT FOR SALE
 ANOVA Table for RCBD
To examine the differences among block effects
as well as among treatment effects, an ANOVA
Table as presented below is utilized
Source of Degrees of Sum of Mean Square F-cal
Variation freedom Squares
Blocks b–1 SSb MSb = SSb/b-1 F = MSb/MSE
Treatments t–1 SSt MSt = SSt/t-1 F = MSt/MSE
Error (b – 1) (t – 1) SSE MSE = SSE/df
Total bt – 1 SST

𝒚𝟐𝒊. 𝒚𝟐. . 𝒚𝟐.𝒋 𝒚𝟐. . 𝒚𝟐. .

𝑺𝑺𝒃 = − 𝑺𝑺𝒕 = − and 𝑺𝑺𝑻 = 𝒚𝟐𝒊𝒋 −
𝒕 𝒏 𝒃 𝒏 𝒏

where SSE = SST – SSb – SSt 𝒂𝒏𝒅 𝒚. . = 𝒚𝒊𝒋

27
 Statement of Hypotheses
The statement of hypotheses to be tested are
similar to those of CRD, except that different
hypotheses are stated to examine the effects of
block and effects of treatments.
Decision Rule
Reject the null hypothesis (H0) if the Fcalculated is
greater than Ftabulated (Fv1, v2, α).
where v1 is the degrees of freedom for block or treatment
as the case may be and v2 is the degrees of freedom for
error
This learning material is NOT FOR SALE 28
 Example 5
In a field experiment to examine the response of maize to
nitrogen fertilizer applications, 6 treatments were used in
4 blocks. The maize grain yields per hectare were
observed and recorded as follows
Blocks
1 2 3 4
1 0.31 0.22 0.33 0.42
Treatments

2 0.58 0.42 0.55 0.46

3 0.83 0.86 0.89 0.60
4 1.68 2.39 2.22 2.63
5 1.95 2.83 2.81 2.81
6 2.50 2.50 2.10 3.39 29
This learning material is NOT FOR SALE
 Example 6
An experiment involved testing the effects of 4 feeds on the number
of eggs laid by pullets kept in each cage. The feeds were randomly
allocated to the cages containing 5 breeds available in Kwara State.
The following table is an abridged ANOVA table for the data
generated from the experiment. Copy and complete the table.
Source of Degree of Sum of Mean Square F
Variation Freedom Squares
Breed 125

Feed 3 56.7

Error 224.8

Total 19

30
This learning material is NOT FOR SALE
 Exercise
An experiment was conducted to compare
the effects on the clotting time of plasma of
four different methods of treatment of
plasma. Samples of plasma from eight
subjects (blocks) were assigned in random
order to the four treatments. Test whether
there is significant difference between the
treatments as well as the subjects at 5%
level of significance. The data recorded for
the experiment are presented below
This learning material is NOT FOR SALE 31
Subjects Treatments
(Blocks) 1 2 3 4
1 8.4 9.4 9.8 12.2
2 12.8 15.2 12.9 14.4
3 9.6 9.1 11.2 9.8
4 9.8 8.8 9.9 12.0
5 8.4 8.2 8.5 8.5
6 8.6 9.9 9.8 10.9
7 8.9 9.0 9.2 10.4
8 7.9 8.1 8.2 10.0
This learning material is NOT FOR SALE 32
#1 Practice
A researcher wishes to see whether there is any difference
in the weight gains of chickens reared in KOK poultry,
each bird was fed with one of the three special diets.
Birds were randomly assigned to three groups and placed
on the diets for six weeks. The weight gains (in kg)
are shown below. At α = 5%, can the researcher conclude
that there is a difference in the diets?
Diet A Diet B Diet C
3 10 8
6 12 3
7 11 2
4 14 5
8
6 33
This learning material is NOT FOR SALE
#2
A researcher wishes to see whether there is a difference
in the weights (in pounds) of four types of lawnmowers.
At α = 10%, can one conclude that the weights differ?
Gas (Self-Propelled) Gas (Push; Rear Bag) Electric Manual
95 73 55 37
101 69 52 24
108 72 51 25
107 71 37 29
97 67 57 22
101 62 54 17
68 34 17
71 45 22
41 20
53 18
21
34
This learning material is NOT FOR SALE
#3

The following data were obtained from an

experiment to test the effects of 3 feeds on the
number of eggs laid by a pen of 6 pullets from each
of 4 breeds. The 3 feeds were randomly allocated to
the 3 cages of each breed. Analyze the data and draw
conclusions
Breeds
1 2 3 4
Feeds

Crumble 330 288 295 313

Mash 372 340 343 341
Pellet 359 337 373 302
35
This learning material is NOT FOR SALE