EXPERIMENTAL

DESIGN

STAB2143
STATISTICAL ANALYSIS IN BIOLOGY

LECTURER: DR. IKMAL

Credit to: Dr. Yosni Bakar

 Two major types of study involved
in research
Experiments vs. Observational Studies

• In an experiment investigators apply

treatments or control the assignment of
treatments to experimental units (people,
animals, plots of land, etc.) and then proceed
to observe the effect of the treatments on the
experimental units.
• In an observational study investigators
observe subjects and measure variables of
interest without assigning treatments to the
subjects.
• The treatment that each subject receives is
determined beyond the control of the
investigator.
 Suppose we want to study the
effect of smoking on lung capacity in women

Experiment Observational Study

• Find 100 women age 20 who do • Find 100 women age 30 of which
not currently smoke. 50 have been smoking a pack a
• Randomly assign 50 of the 100 day for 10 years while the other
women to the smoking treatment 50 have been smoke free for 10
and the other 50 to the no years.
smoking treatment. • Measure lung capacity for each of
• Those in the smoking group the 100 women.
smoke a pack a day for 10 years • Analyze, interpret, and draw
while those in the control group conclusions from data
remain smoke free for 10 years.
• Measure lung capacity for each of
the 100 women.
• Analyze, interpret, and draw
conclusions from data.
 Why Experimental Design?

• Malay Proverb
‘Bagaimana Acuan Begitulah Kuihnya’
The (shape of a)cake reflects its mold
 Why experimental Design?
• Phases of every project
Statement of problem
Hypothesis
THE EXPERIMENT What to measure?
Define factors

Sample size (how

How to carry out experiment so
large a difference is
data collected is
to be significant) THE DESIGN
-relevant to the problem.
Method of
randomization -Permits objective analysis.
Statistical model -Valid inferences to the problem

Data processing
Computation of test
THE ANALYSIS
statistics and
Interpretation
 Terminology
• Study the effects of water temperature and level of
UV radiation on the growth of fish
§ Temperature
• 25 C and 30 C
§ UV
• Low and High

Low UV
25 C High UV

30 C Low UV
High UV
25 C 30 C 30 C 25 C
LOW UV LOW UV HIGH UV LOW UV

25 C 30 C 25 C 30 C
HIGH UV HIGH UV LOW UV LOW UV

25 C 30 C 30 C 25 C
HIGH UV HIGH UV LOW UV HIGH UV
 Terminology
FACTORS
LEVELS TREATMENTS
Variables under the
control of the Values of the factors combination of
experimenter factor levels
applied to an
Also called explanatory
or independent experimental
variables 25 C unit
A factor has 2 or more
levels
30 C
TEMP TREATMENTS

HIGH
UV
LOW
 Terminology
• What is being measured
Response • Also called dependent variable
variable • e.g. Body weight of fish after 30
days

• The unit to which the treatment

is applied
Experimental • e.g. The tanks are exposed to
unit different treatments,
• So tanks are experimental units.

• The unit on which the response

is measured i.e. fish.
Observational • Caution: May not be the same as
unit experimental unit as shown
here.
A slightly different case in which each tank has only
one fish. Here, the weight of the fish in a tank is the
observational unit for the tank, meaning the
experimental unit is the same as the observational unit.

25 C, Low 30 C, High

30 C, Low 25 C, High

25 C, Low
 The difference:
when
experimental unit ≠observational unit experimental unit = observational unit

Suppose
Treatment: 4 Treatment: 4
Replicate or tanks for 3 replicates per treatment
each treatment: 3 1 fish to a tank
2 fish to a tank
Sources of variation
Sources of variation Treatments
Treatments Replicates (treatment)
Replicates (treatment)
Fish (replicate, treatment)
 Example: Terminology
• MARDI is going to test two
varieties of pineapple. Each Identify:
variety will be tested with two 1. Factor(s)
types of fertilizers. Each
combination will be applied to 2. Level(s) of
two plots of land. The yield will each factor
be measured for each plot.
3. Experimental
units
• Dr Fauzi is testing the effect of
4 substrates on the growth of 4. Observational
mushroom in petri dish. Each units
substrate is replicated in 4 5. Response
petri dishes. The average variable
length of mycelium in a petri
dish will be measured.
 Key considerations in designing an
experiment

• Simplicity: choose the simplest design

among the many designs that can
achieve the same objectives.

• Efficiency: Save time, money, personnel

and experimental materials.
Basic principles of experimental
design (3 R’s)
• The basic objective of
experimental design is to reduce
the experimental error
A C A B

• What is experimental error?

Experimental error – a measure
of variation among observations
on experimental units receiving
the same treatment i.e. when A B C C B
units fail to yield identical results

Growth of seedlings applied with 3

• How can we estimate and reduce different brands (A,B,C) of fertilizer
the experimental error?
3 R’s Growth of plants applied with different
fertilizers (A,B,C), each in 3 replicates or
plots

A
B
C

1. Provides means of
estimating experimental
Replication. error (no replicate, no way
What is it? to estimate error)
Repetition of treatments in 2. Improve precision of
an experiment. estimate of treatment
means. S.E mean =
√(1/n)σ2, σ2 is
experimental error, n is
number of replicates
 Growth of plants applied with different
fertilizers (A,B,C), each in 3 replicates or
3 R’s plots

Though the layout of the

experiment allows
estimate of experimental A
error, is it an unbiased
B
and valid estimate? NO !
C

RANDOMIZE

Randomization
What is it?
Random allocation of treatments to
the experimental units.
Why?
Randomization tends to average out
the influence of extraneous factors not
under the control of experimenter
Data free from any systematic error
3 R’s

StRatify or blocking
 Structures of an experimental design
Consists of two basic structures
• How the experimental units
are arranged.
Treatment structure
• Are they being grouped into
homogeneous groups or blocks? • How the treatments are
arranged or combined in the
• Common design structures are: design structure.
• Completely randomized • consists of a set of treatments
design (CRD) e.g. A set of different
• Randomized complete block Concentrations (one-way
design (RCBD) treatment structure)
• Latin squares design (LSD) • Or a set of treatment
combinations e.g.
Concentrations and
Temperature (two-way or
Design structure multi-way factorial, split plot
or nested arrangement)
• Once the treatment and design structures are
selected, need to randomize the treatments of the
treatment structure TO THE experimental units of the
design structure

TREATMENT
DESIGN
STRUCTURE
STRUCTURE

RANDOMIZATION

So, designing an experiment involves:

a) choice of treatment structure
b) choice of design structure
c) method of randomization
Overview
DESIGN
STRUCTURE

Completely Randomized Complete

Latin Squares
Randomized Design Block Design

One Factor Two or more Two or more

One factor One factor
Analysis factors factors

Factorial Factorial

TREATMENT
nested STRUCTURE Split-plot

Split-plot