I

RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

RESEARCH DESIGN
• Logic of the inquiry
• Purpose of the inquiry
• Types of research design
• Research study (design) vs research methodology
• Direction of the reasoning

Yos Adi Prakoso, DVM., M.Sc - Research Design I

BASIC CONCEPT
• Research methodology
Quantitative, qualitative
• Research methods
Methods of sampling, data collection and analysis
• Research design
Experimental, descriptive, exploratory

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 THE BASIC FRAMEWORK
Data

Results

Methodological designs Discussion

Choice of research design

(Types)

Statement of research problems

Unit of analysis and objectives
The logic of the research (reasoning and conceptualising)
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 PURPOSE OF THE INQUIRY
Exploration

Description

Explanation

Prediction

Evaluation

History Yos Adi Prakoso, DVM., M.Sc - Research Design I

 EXPLORATORY RESEARCH
Aims Most useful when the problem is not
well understood
High degree of flexibility

To identify the salient factors or

causal links
Referred to analytical study

Very structure in nature

Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DESCRIPTIVE RESEARCH
Aims Provide accurate information

Valid representation (encapsulated)

factors or variables

More structured than explanatory

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 RESEARCH DESIGN VS RESEARCH
METHODOLOGY
Research
Research Design
Methodology

Focus on end-product Focus on research process

Point of departure: specific

Point of departure: research
task (data collection and
question and problems
sampling) at hands

Focus on the logic research

Focus on the steps and
and what evidence is
required procedure to be employed
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DIRECTION OF THE REASONING

Patterns Theory

Deduction
Induction

Observations/
Data Hypothesis

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 COMPARISON OF DEDUCTIVE AND
INDUCTIVE
Hypothetic- Ethnographic-
deductive logic inductive logic

Gain experience,
Develop an idea
participate, listen, records
(theoretical framework)
experiences, and data

Describes the theoretical

Test it implication (discussion
and theory building)

See if you were right or not

Explain
(discussions)
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 TYPES OF RESEARCH DESIGN
Generating Primary Analysing Existing
Data Data

Text data: discourse and

Survey, case study,
content analysis, textual
program evaluation
criticism, historical studies

Numeric data: secondary

Experiment data analysis, statistical
modelling

Ethnographic study
Yos Adi Prakoso, DVM., M.Sc - Research Design I
II
RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

 QUANTITATIVE RESEARCH
The best way to investigate the structure
not process
The specific question is how, what and
where
Predetermined responses of categories
and data collection
Enhance the validity and reliability of
observation (counting)
To facilitate replication studies

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 SAMPLING IN QUANTITATIVE
RESEARCH
Aims Generalisation to a larger (study)
population

Sample size (large), random

sampling, why?

Verification of theory

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 DATA ANALYSIS IN QUANTITATIVE
RESEARCH
Descriptive Inferential
Statistic Statistic
• Simple distribution (one • Use probability theory:
variable ) • To test hypothesis
• Bivariate relationship (two • To draw inferences as
variables: frequency to whether results
distribution) • To test whether
• More than two variables descriptive results
(tri/multivariate
variables: multiple
regression analysis)
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 NEED TO KNOW: CONCEPTS
Statistical Significance
• Means that’s results to be due to chance factors
• It tells the researcher that the results produced
by the random error in random sampling
• Results can be statistically significant
theoretically meaningless or trivial
Probability Theory
• Refers to process that generate a mathematically
random results
• The researcher can calculate the probability of
outcomes Yos Adi Prakoso, DVM., M.Sc - Research Design I
 STEPS IN DESIGNING A
QUANTITATIVE STUDY
Formulate a researchable question

Review related literature

State hypotheses

Determine the variable to be studied

Determine the research plan/method of data

collection
Define population

Determine what instrument will be used to collect

the data
Determine statistical test to use Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DATA COLLECTION IN QUANTITATIVE
STUDIES
Experimental
• Simple post test
• Classic pre-test and post-test
• Pre-test, post-test, control group
Secondary Analysis of Quantitative Data

Observation
• Use check or tally sheet

Surveys
• Use questionaries
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 QUALITATIVE RESEARCH
To get close to the data in their natural
setting

To reflects an individual’s experience in

the context of their everyday life

Uses small sample size and dig s deeply

for data
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 SAMPLING IN QUALITATIVE
RESEARCH
Sampling is mostly purposive
Aims
Seek conceptual applicability rather than
representativeness (quantitative
representivity)
You want to capture the ranges of view

Seek after/pursue saturation of data

Draw theory from data

Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DATA COLLECTION IN QUALITATIVE
STUDIES

Participant Observation

Case Studies

Formal and Informal Interviewing

Videotaping

Archival Data Survey of Document Review

Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DATA COLLECTION IN QUALITATIVE
STUDIES (CONTINUATION…)
Emphasises comprehensive, inter-independent,
dynamic and holistic structure

Appropriate in the investigation of the messy

problem and complex

Often draws multiple sources of data

Given its strength in generating/ developing theory

(inductive)
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 DATA ANALYSIS IN QUALITATIVE
STUDIES
Discourse analysis

Narrative analysis

Content analysis

Thematic analysis
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 TRIANGULATION IN QUALITATIVE
RESEARCH
Data Triangulation
• Multiple data to understand a phenomenon

Methods Triangulation
• Multiple research method to study a phenomenon

Researcher Triangulation
• Multiple investigator in analyzing and interpreting the data

Theory Triangulation
• Multiple theories and perspective to help interpret and
explain the data
Yos Adi Prakoso, DVM., M.Sc - Research Design I
III
RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

 HOW TO SOLVES THE PROBLEM IN
THE DATA ANALYSIS
Arrange the Simple Data
• Make a summary of data univariate

Draw the Relation between Data

• Find the relationship and correlation of data

Make a decision
• Decision needed to draw the conclusion
remarks Yos Adi Prakoso, DVM., M.Sc - Research Design I
 HOW TO SOLVES THE PROBLEM IN
THE DATA ANALYSIS
Arrange the Simple Data
• Make a summary of data univariate

Draw the Relation between Data

• Find the relationship and correlation of data

Make a decision
• Decision needed to draw the conclusion
remarks Yos Adi Prakoso, DVM., M.Sc - Research Design I
IV
RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

 COMPLETELY RANDOMIZED DESIGN
(CRD)
Simplest design set-up

Easiest to do

Easiest to analyse

Often sufficient for the goal

Treatment are randomly assigned to …

Yos Adi Prakoso, DVM., M.Sc - Research Design I
 COMPLETELY RANDOMIZED DESIGN
(CRD)
Has N = units
g = group with different treatment
ni = number of the observation in
each treatment, the formulae below:

෍ 𝒏𝒊 = 𝑵

Completely random assignment of

treatments to unit
Each group has a equally sample
size Yos Adi Prakoso, DVM., M.Sc - Research Design I
 COMPLETELY RANDOMIZED DESIGN
(CRD)
F Controlled
We have ability to choose and set
A Ex: age, sex
C Uncontrolled
Have a minor effect
T
Primary factor
O Factor of interest
R Nuisance factor
The factor may include in the research but not
primary interest
In this case, it also called as
Yos Adi block
Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE 1.A
One-Factor Study CRD
We have 8 chickens fed a diet naturally
contaminated with aflatoxin. The chickens were
separated into 2 groups. The histopathological
change observed in day 5 (group 1) and 10 (group
2) post treatment.
• N = 8 chickens
• g = 2 groups
• n1 = n2 = 4 chickens
• To make it easy the table prepares as follow:
Group 1 Group 2

4 chickens 4 chickens
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE 1.B
One-Factor Study CRD
This was the CRD study assigned to 12 marmots as the
models for cardiomyopathy. The marmots treated with
0, 10, 20, 40 mg/kg BW of X drug. The parameter was the
increasing level of cardio-tension.
• N = 12 marmots
• g = 4 groups
• n1 = n2 = n3 = n4 = 3 marmots
• To make it easy the table prepares as follow:

Doses Cardio-tension
0 80 80 85
10 85 85 85
20 100 100 102
40 125 125 120
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 In the Single Factor Experiment, the
Usual Items of Interest …
Is there evidence that some means are different?

If they are different, which are different from the

other?

Any pattern in differences?

Estimates/ confidence intervals of means and

differences

In some cases, variability maybe of interest

Yos Adi Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE 2
Two or More-Factor Study CRD
We have 8 chickens fed a diet naturally contaminated with
aflatoxin. The chickens were separated into 2 groups. The
histopathological change observed in day 5 (group 1) and 10
(group 2) post treatment. Besides, histopathology the research
observed the two level of doses of aflatoxin.
• N = 8 chickens
• g = 4 groups
• n1 = n2 = n3 = n4 = 2 marmots
• Factor = day and doses
• To make it easy the table prepares as follow:
DAY LENGTH (5 and 10)
DOSES (1 and 2)

A B
C D
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE 2
Two or More-Factor Study CRD, continuation…
Those factor could also perceived as a single variable

It called as super-factor

There is no block and nesting

Doses 1 Dose 1 Dose 2 Dose 2

Day 5 Day 10 Day 5 Day 10

Yos Adi Prakoso, DVM., M.Sc - Research Design I

V
RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

 RANDOMIZED COMPLETE BLOCK
DESIGN (RCBD)
A block of units is a set of units that are
homogeneous in some sense

To form blocks, we organize EUs into groups having

similar characteristics
For example: age, family, location, etc

Remember, don't use a primary factor of interest for

blocking

A blocking factor should be a nuisance factor.

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 RANDOMIZED COMPLETE BLOCK
DESIGN (RCBD)
Other ways to form a block (ideally, we want to be
able to observe all treatments within each block):
For example:
1. Physically divide an object into parts, like a
manufacturing setting
2. Repeat testing of the same object under the
diferent conditions (like when you give both drugs
to the same person).
A RCBD utilizes restricted randomization, where we
randomly assign the treatments to the EUs within a
block. Thus, if there are r blocks, then we will do r
restricted randomizations, one for each block, to assign
the treatments to the EUs.

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 RANDOMIZED COMPLETE BLOCK
DESIGN (RCBD)
In a complete block design, all treatments are
observed on each block. If you can not observe all
treatments on each block, then you have an
incomplete block design.
Blocked designs are not CRDs

Blocking is a variance reduction technique

Block-to-block variability is still in the data, but we

essentially remove this variability when comparing
treatments (because we see all treatments within a
Yos Adi Prakoso, DVM., M.Sc - Research Design I
block)
 RANDOMIZED COMPLETE BLOCK
DESIGN (RCBD)
Blocking is most useful when there is wide variability
across blocks.

We don't usually test for a block effect because we

EXPECT a large difference across blocks, that's exactly
why we're using it, and it's a nuisance factor anyway

In general, we assume there is no interaction between the

block and the treatment. We assume the treatment effect
(i.e. differences between the treatments) is the same for
all blocks
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE 1.B
One-Factor Study CRD
Drugs A, B, C, and D are to be compared. We have formed blocks based on
age, and we have 4 patients in each of 6 age blocks. We are not interested
in testing for an age effect (it is a nuisance factor) as we are most
interested in comparing treatments. Within each block, we randomly assign
a unique drug to the 4 patients.

Ages

1 2 3 4 5 6
Drugs

A
B
C
D Yos Adi Prakoso, DVM., M.Sc - Research Design I
 By comparing the block so that
RCBD…
Uses `restricted randomization', performed within each block.
* g treatments
* g EUs per block
* r blocks
* rg = N total units

It's like r single-replication CRDs glues together.

The RCBD is used to increase power and precision of

an experiment by decreasing the error variance used in
testing
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 RANDOMIZED COMPLETE BLOCK
DESIGN (RCBD)
By comparing treatments within a block, we
remove the block-to-block variability from our
treatment comparison analysis

Blocking is a powerful tool and should be used

if possible to control for any `nuisance'
variation that is thought to be large.

Yos Adi Prakoso, DVM., M.Sc - Research Design I

VI
RESEARCH DESIGN

Yos Adi Prakoso, DVM, M.Sc.

 FACTORIAL DESIGN
Most efficient to evaluate the effects of two or more
factors

All possible combination of the levels of the factors

are investigated in each replication

If there are a level of A factor, and b level of B factor,

then each replicate contains all ab treatment
combination

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 FACTORIAL DESIGN
M The main effects of a factor is difined to
A be changed in response produce by a
change in the level of a factor
I
N
The main effects of A is the difference
between the average response at A1 and
E A2
F FACTOR B
F
A1, B1 A1, B2
FACTOR A

E
C
A2, B1 A2, B2
T
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 FACTORIAL DESIGN
I In some experiment, we may find that the
N difference in response between the level
of one factor is not the same at all level of
T the other factor
E
If it occur, there is the interaction between
R factors
A
C FACTOR B
T
A1, B1 A1, B2
FACTOR A

I
O
A2, B1 A2, B2
N
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 INTERACTION GRAPH OF THE
FACTORIAL DESIGN

60
A factorial experiment
50
Response

40 without interaction
30
20
10
0

- +
Factor A
B- B+

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 INTERACTION GRAPH OF THE
FACTORIAL DESIGN

60
A factorial experiment with
50
Response

40 interaction
30
20
10
0

- +
Factor A
B- B+

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 FACTORIAL DESIGN
N Graph are frequently useful in interpreting
O significant interaction
T
E
When an interaction is large, the main
effects have little practical meaning

A significant interaction will often mask

the significant of the main effects

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 FACTORIAL DESIGN
A They more efficient than one-factor-at-a-
D time experiment
V
A
N When the interaction may be present, it
can be useful to avoid misleading
T conclusions
A
G
Factorial design allow the effects of a
E factor to be estimated at several levels of
S the other factors, conclusions are valid
over a range of experimental condition
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 TWO-FACTORIAL DESIGN
The simplest type of factorial designs involve only
two factors or sets of treatments

There are a levels of factor A, and b levels of factor

B, and each replicate contains all ab treatment
combinations

In general, there are n replicates

Yos Adi Prakoso, DVM., M.Sc - Research Design I

 EXAMPLE
Two-Factorial Design
An researcher test the weight increasing of 3 species
of fish in three different temperature of the pond (15o F,
45o F, and 60o F). Three fish (replicates) are tested of
each combination of species and pond temperature,
and all evaluated 3 times in random order.

• Question:
• What effects do species and temperature
have on the life of a battery?
• Is there a species that would give long life
regardless of temperature?
Yos Adi Prakoso, DVM., M.Sc - Research Design I
 EXAMPLE
Two-Factorial Design
Data

Species Ponds temperature

Total
15o F 45o F 60o F
130 34 20
1 74 359 80 154 70 172 685
155 40 82
150 136 25
2 159 497 122 364 70 153 1.014
188 106 58
138 174 96
3 168 416 120 444 104 282 1.142
110 150 82
Total 1.272 962Yos Adi Prakoso, DVM., M.Sc607
- Research Design I
 INTERACTION GRAPH

600
Two-factorial experiment
Weight increasing

500
400 with interaction
300
200
100
0

15 45 60
Temperature
species 1 Species 2 Species 3

Yos Adi Prakoso, DVM., M.Sc - Research Design I