Dr. Mazharul Islam

What is research design?

A research design is the arrangement of conditions for collection and analysis of data in a manner that aims to combine relevance to the research purpose with economy in procedure.

Properties or Characteristics of a Good Research Design

In the view of various definition of research design, the following characteristics are found.

•    A good research design is an ethical research design;

•   A good research design is one that is capable of obtaining the most reliable and valid data;

•  A good research design is one that is capable of measuring any odd events in any circumstances;

•  A good research design is one that helps an investigator avoid making mistaken conclusions;

•  A good research design is one that can adequately control the various threats of validity, both internal and external.

Need of Research Design: Research design is needed because it facilitates the smooth sailing of the various research operations, thereby making research as efficient as possible yielding maximal information with minimal expenditure of effort, time and money. Research design stands for advance planning of the method too be adopted for collecting the relevant data and the techniques to be used in their analysis, keeping in view the objective of the research and the availability of staff, time and money.

Different Types of research design

There are two types of research design such as

1. Experimental Research Design and

2. Non-experimental Research Design

1. Experimental Design: Experimental research design is an elegant way to find out how well a particular program achieves its goals. In which, the researcher       manipulates the situation and measures the outcome of his manipulation.

2. Non-experimental Design: A non-experimental study is one in which the researcher just describes and analyze researchable problems without any manipulation of the situation.

Types of Non-experimental Design:

There are three types of non-experimental design such as:

I. Exploratory studies;

II. Descriptive studies; and

III. Causal studies.

Exploratory Studies: An exploratory study is a small scale study of relatively short duration which is undertaken when little is known about a situation or problem.

Types of Exploratory Study: Exploratory study offers an opportunity to obtain insights into the problem through four categories such as i. Secondary Data Analysis; ii. Experience Survey; iii. Pilot Study; and iv. Case Study.

Descriptive Studies: Descriptive studies are those used to describe the characteristics of a population or phenomena. The objective of the descriptive study is to focus on “who”, “what’, “when” and “how” question.    

Types of Descriptive Studies: There are seven types of descriptive studies such as i. Cross-sectional Study; ii. Longitudinal Study; iii. Trend Study; iv. Panel Study; v. Baseline Study; vi. Impact Assessment Study; and vii. Feasibility Study.

Causal studies: A casual also called explanatory or analytical study attempts to establish cause or risk factors for certain problems. Our concern in causal studies is to examine how one variable “affects “or is responsible for changes in another variable.

Types of Descriptive Studies: There are three types of causal studies such as

i. Comparative Study;

ii. Case-control Study; and

iii. Cohort Study.

Experimental Design: Experimental research design is an elegant way to find out how well a particular program achieves its goals. In which, the researcher manipulates the situation and measures the outcome of his manipulation.

Characteristics of Experimental Design: The classical experimental study has three characteristics such as a) Manipulation; b) Control; and c) Randomization.

Manipulation:  Individuals, in an experimental study are randomly selected and allocated to at least two groups. One group is subject to an intervention, often called experimental or test stimulus, while the other group(s) is not. In a true experimental study, the experimenter has the ability to measure the values of the dependent variable both before administering the stimulus and after administering it. The difference between these scores gives a rough indication of the effect of the causal variable.

Control: By control, we mean, all factors with the exception of the independent variable must be held constant and not confounded with another variable, that is not part of the study.

Randomization: By randomization, we mean that the researcher takes care to randomly assign subjects to the control and experimental groups, meaning that each subject is given an equal chance of being assigned to either group.

Completely Randomized Design

To Set up a Completely Randomized Design (CRD) for 3 treatments T₁,T₂,T₃  with 4 replications we have to consider the treatments as T₁,T₁,T₁,T₁,T₂,T₂,T₂,T₂, T₃, T₃,T₃,T_3. 

Now we have to allocate these 12 treatments randomly in 12 plots using random numbers as follows

Yield or Outcome

Notation

y_ij= observation j in group I;  i=1, 2, 3, . . . ,k=4 and j=1,2,3,…,n_i ; N= Σn_i=16

ӯ_i. = mean of data in group i

y_i.= total of data in group i

ӯ_..= mean of all data

y_..= total of all data

The partitioning of sums of squares corresponds to partitioning of differences:

Deviation of observation from overall mean= (deviation of observation from group mean) + (deviation for group mean from overall mean)

                                                y_ij −ӯ_..=  (y_ij− ӯ_i. ) + (ӯ_i. − ӯ_.. )

Sum square of deviation of observation from overall mean = (Sum square of deviation of observation from group mean) + (Sum square of deviation for group mean from overall mean)

                          Σ Σ(y_ij −ӯ_..)² = Σ Σ(y_ij− ӯ_i.)² + Σ Σ(ӯ_i. − ӯ_..)²

                              Σ Σy_ij ²- (y_..²/N)= Σ Σ(y_ij− ӯ_i.)² + Σ(y_i.²/n_i) - (y_..²/N)

                          SS(Tot) =   SS(Err)     +         SS(Tr)

Analysis of Variance (ANOVA)

Let us consider 5 Treatments: T1,T2,T3,T4 & T5 and No. of blocks: 4
Random selection of Treatments: 

RANDOMIZED BLOCK DESIGN ( RBD)

Yield or Outcome

Notation:

yij= observation of jth treatment and ith block;

ӯi. = mean of ith block;          yi.= total of ith block

ӯ.j = mean of jth treatment;    y.j= total of jth treatment

ӯ..= mean of all data

y..= total of all data

j=1, 2, 3, . . . ,k=5 and i=1,2,3,…,n=4 ; N= nk=20

Sum square of deviation of observation from overall mean= (Sum square of deviation for block mean from overall mean) + (Sum square of deviation for treatment mean from overall mean) + (Sum square of deviation of observation from block mean and treatment mean)  

  Σ Σ(yij − ӯ..)2 = Σ Σ(ӯi. − ӯ..)2+Σ Σ(ӯ.j − ӯ..)2+ Σ Σ(yij − ӯi.- ӯ.j +ӯ..)2  

  Σ Σyij 2- (y..2 /N)= Σ Σ(yij − ӯi.- ӯ.j +ӯ..)2 + Σ(yi.2/k) - (y..2 /N) + Σ(y.j2/n) - (y..2 /N)

  SS(Tot) =   SS(Err)     +  SS(blok)    +    SS(Tr)

Analysis of Variance (ANOVA)Table:

Latin Square Design

ii) Randomize the rows of selected Latin square

RANDOM ROWS:  R5,  R3,  R2, R4,  R1

Contributor: Dr. Mazharul Islam is a univrsity teacher working as an Assistant Professor, Department of Business Administration, BIU, Dhaka, Bangladesh.  He has done his Ph.D from Rajshahi University in Statistics.