Module 4: DESIGNING THE RESEARCH METHODOLOGY

Learning Objectives
At the end of this chapter, the reader should be able to:
 discuss the different types of research design
 differentiate the types of sampling design
 explain the different method of data collection and data analysis

4.1 Research Designs

Research experts have different views about the definition of research design. Based on these
views, research design appears to be a scheme that maps out the source of data, the type of data
to be collected, how the data will be collected, and the methods to be used in data analysis. The
plan also pinpoints specifically the relationship of the variables under study. A good research
design must also set time constraints within which the research problem should be answered.

A multifaceted concept, research design may be classified in terms of the purpose of the
study, timeframe, scope of the problem, data requirement, and method of data collection. The
type of business research design includes descriptive, exploratory, causal/explanatory,
correlational, comparative, evaluative, experimental, and action research. An undertaking may
adopt more than one research design.

1. Descriptive Research is conducted to illustrate and determine patterns or characteristics of

variables in particular instances or events. In the statement of specific objectives some of the
 more appropriate verbs to use are describe, determine, analyze, explain, define, and
illustrate.

The study entitled “Perceived Strength and Weaknesses of Solidbank Corporation’s Job
Rotation Program” used a descriptive research design as presented below:

The study utilized a descriptive research design, the purpose f which was to capture the
relevant issues that surround job rotation. Its other purpose was to obtain the respondents’
perceived strengths and weaknesses of the job rotation program with regard to promotion
opportunities, opportunity for personal growth, recognition, responsibility, achievement,
quality of supervision, pay, company policies, physical working conditions, relation with
others, and job security.
A five-point Likert scale survey questionnaire was used to obtain the data.

The research design of the investigation entitled “An In-depth Study of the Situation of
Women Trade Unionists in the Philippines” is described as follows:

The research problem and objectives posed at the beginning of the study were answered
through a descriptive research design. The design focused on describing the women
union workers’ personal profile, profile on family, community, work, politics and union-
related data.
In the women workers’ profile, analysis of patterns to describe their age, civil status,
educational attainment, average monthly income and their socio-economic conditions
was made.
In their family-related profile, roles, decision-making and family relationships were
analyzed. Respondents also cited the difficulties encountered in portraying both family
and work roles through an open-ended statement.
With their work-related data, there was an analysis to describe workers’ length of
service in the company, number of promotions, and ways of getting hired in the
companies. The existence of discrimination in their company was also discussed.
With their union-related data, there were open-ended questions related to their positions
in the union, number of years in the union, ways of participation, and their level of
assessment of the union and their union leaders.

2. Exploratory Research is undertaken to gather initial data patterns or characteristics of

variables in a situation where studies have been done or minimal data are available to
establish significant patterns or relationships of variables in a specified scope. In this
particular design, the statement of hypothesis is not usually present because one of the
purpose s of an exploratory study is to recommend possible hypothesis which can be tested
in future studies. In the statement of specific objectives, some of the more appropriate verbs
to use are explore, find out, know, examine, and investigate.

In the study that seeks to find out the applicability of virtual offices in Philippine companies,
the research design can be presented as follows:

The study made use of exploratory research to find out the potential of having virtual
offices in Philippine companies. Virtual offices are being employed in other countries by
a few companies. Some articles have also been written about the advantages and
disadvantages of implementing such setup. However, no research has yet been made on
 its applicability or on the perceptions of executives regarding the matter. The proponents
used exploratory research to narrow the scope of the research topic into defined ones. It
was not intended to provide conclusive evidence, but rather to examine it further so as to
provide directions for future research.

Another exploratory research design on Green Marketing can be stated this way:

The extent by which marketing executives understand “Green Marketing” has not been
examined as much in studies made. That is the reason why the exploratory research
design was employed for this study. The study is geared towards investigating the
concept of green marketing as understood and practiced by marketing executives in their
companies.

3. Causal/Explanatory Research is carried out to ascertain that the occurrence of, or change
in, the independent variables. In the statement of specific objectives, some of the appropriate
verbs to use are analyze, explain, and identify.

The study entitled “Product Optimization for the Philippine Tourism Industry” made use
of the causal research design as shown below:

The plan and structure of the investigation were conceived along marketing research’s
concept testing procedure which relies on concept evaluation and positioning by market
segments.
The investigation was divided into two parts–the perceptual side and the choice of the
total tourist product.
Since the focus of part one (demand) is accurate description of variables and
specification of a functional relationship between them, descriptive research was utilized.
In part two, it was necessary to use the causal approach because, in product
conceptualization, the main concern is knowing how one variable affects or is responsible
for the changes in another variable. To understand such influences, it is necessary to
manipulate or change the variables representing attributes of interest and observe the
concomitant variation in the preference.

Another study that made use of the causal research design is entitled “The Effects of Quality
Circles in Company E”, as presented below:

The research used is the causal method which was pertinent in gathering salient data on
the effects of quality control as a business management strategy applied in company E.
The causal method takes into account all important aspects of employing such business
strategy technique to determine its effects. The researcher used the causal method
because it probes deeply and analyzes interaction between the manager and the workers.
In addition to that, the study employed various data probing technique to arrive at an
impartial assessment and understanding of the study.

4. Correlational research is conducted to establish the fact that the outcome of certain
patterns of relationship occur together in a specified manner without suggesting that one
variable causes the other variable causes the other variable to change. In this design,
relationship is also clearly identified as positive or negative correlation. The most
appropriate word to use in starting specific objectives is correlative or interrelate.
 The study entitled “Chief Executive Scanning, Environmental Characteristics and Financial
Performance: A Canonical Analysis and Empirical Investigation of Top United States
Manufacturing Firms in the Philippines” presented its design as follows:

The correlation design was used for two reasons. One was to establish the degree that
the indicator variables are associated to one another, to attest whether the theorized
autonomy of these variables is supported by the data. And the other reason was to
describe in quantitative terms the extent to which the scanning variables were related to
one another, both to test hypothesis and to determine the make-up of the ensuing analysis.

Correlational research design is used in the study, “The Exchange Rate Pass-Through in the
Philippine Manufacturing Industries: A Cross-Sectional Analysis of the Magnitude of Exchange
Rate Elasticities with the Degree of Product Differentiation, Elasticity of Marginal Cost and
Market Share ”, as presented below:

The study employed the correlational research design with the use of two-stage
regression to answer the research problems and objectives posed at the beginning of the
study.
The first stage regression focused on determining the correlation between the monthly
movement of the exchange rate and the variation on the producer’s price index per
manufacturing industry. A time-series regression analysis was used to get the significant
coefficient for the slope for each exchange rate pass-through to be used as variable for the
second stage regression.
On the other hand, the second stage regression used a cross-sectional analysis between
the exchange rate pass-through (coefficient of ER from the first stage regression) against
the three parametric models, namely: the degree of the product differentiation, the
elasticity of marginal cost and the foreign firm’s share in the industry. The second-stage
regression mainly delved on measuring the magnitude of the pass-through by setting the
responsiveness of each parameter to the pass-through.

5. Comparative Research is undertaken to confirm if two or more variables reveal similar or

different patterns of characteristics when compared, using a set of variables as standard
bases. In the statement of specific objectives, some of the more appropriate verbs and
phrases to use are compare, determine similarities, find out differences, and contrast.

The paper, “A comparative Study Between the Japanese Principles of Personnel Management
and the Actual Practice in a Philippine-Based Japanese Company” presented its research design
as follows:

The study the researchers made was basically descriptive in nature, and more
specifically, a comparative one. In this context, the researchers’ aim was to compare the
principles of Japanese Personnel Management with that of Pilipinas Nissan Company
Incorporated. The researchers agreed that among the various Japanese car-manufacturing
companies here in the Philippines, Pilipinas Nissan Inc. would serve as the basis of the
said comparison.
The point of comparison would be the features of the Japanese way of personnel
management such as recruitment and hiring, training, job description, promotion and
advancement, and duration of employment.
 The paper “Human Resources Management Practices of Small, Medium and Large
Enterprises in Metro Manila” employed the descriptive and comparative designs, as shown
below:

The descriptive and comparative research designs were employed to answer the research
problems and objectives posed at the beginning of the study.
The descriptive design focused on describing the trends of the sample companies’
HRM-implemented policies, programs and activities related to procurement,
compensation, development, integration, maintenance and separation. There was also an
analysis of patterns to describe the characteristics of the companies such as years of
operation, nature of business, size ownership, unionization, and level of profit.
On the other hand, the comparative design delved on determining the differences or
similarities in the enterprises’ HRM practices using their company characteristics as the
basis of comparison. The comparative analysis also attempted to identify significant
correlations of the variables under study.

The paper, “An Assessment of the Business Management Program of the DLSU as perceived
by the Business Management Graduates of School Year 1981–1985” used both descriptive and
comparative designs, as presented below:

The study made use of the descriptive and, partly comparative research designs. Trend
and patterns of BM graduates’ perception of the BM program were presented and
analyzed and these initial findings were compared between pure Commerce-BM
graduates and LIACOM-BM graduates.
The feedback of companies that hire DLSU BM graduates were also shown.

6. Evaluative Research is conducted to assess performance outcome or impact of a set of

variables on another one. This design considers an adequate time span to pinpoint the effects
accurately. This kind of research could also be a longitudinal study. The appropriate verbs to
use in the statement of specific objectives are evaluate, assess, appraise, measure, and
quantify.

Illustrative of the evaluative research design is the paper entitled “An Assessment of the
Skills Needed by the Finance Sector in Relation to the Management of Financial Institution
Curriculum of De La Salle University”, as shown next page:

Currently, providing quality service to customers is the number one mode of fighting
the stiff competition in the business. This study was done to evaluate Kodak’s Delivery
service. The different factors that were discussed under this study are the timing of
delivery, condition of the product, accuracy and documentation of the products,
completeness, and attitude of the delivery personnel. These factors were evaluated to
determine the strength and weaknesses of the company’s delivery service. This evaluative
research design was utilized to investigate potential opportunities or problems.

The study “Reflections on the Road Travelled – A study on the Gender Program of LEARN”
also made use the descriptive and evaluative research designs, as shown below:
 The descriptive and evaluative research designs were employed to answer the research
objectives posed at the beginning of the study.
The descriptive design focused on illustrating the profile of the respondents, CBA
provisions on gender concerns, and the development of the LEARN gender program
during the past 6 years.
On the other hand, the evaluative design delved on determining the effects/impact of
the gender education program both on the individual participants and on their unions.

7. Experimental Research is undertaken to determine the possible cause-and-effect

relationship through experimental and control groups. Such research design is usually done
in a simulated environment. The experimental group, which maybe one or more, is exposed
to the treatment conditions or manipulations, while the control group, which may also be
one or more, does not receive the treatment and is used to compare the results. It is also
possible under limited conditions to conduct actual field experiments to approximate the
conditions of the true experiment. This design is called quasi-experimental. Due to its
limitations, its internal and external validity may be open to questions. In the statement of
specific objectives, some of the more appropriate phrases to use are determine the effects,
examine the differences, compare the results, and analyze the influence.

The excerpt that follows was taken from the paper, “Interpersonal Relationship Skills
Building Program: An Experimental Study.”

This study used the experimental method of research to assess the effectiveness of the
Interpersonal Relationship Skills and Teambuilding Program conducted by the company.
The experimental design was the pretest-posttest control group design which takes the
following form:

(E) R O1 X O2
(C) R O3 O4

where:
R = random selection and distribution of subjects
E = experimental group having treatment
C = control group having no treatment
O1 = Interpersonal Relationship Skills Inventory (IRSI)
O3 = Interpersonal Communication Inventory (ICI)
X = treatment (Interpersonal Relationship Skills and Teambuilding Program)
O2 = Interpersonal Relationship Skills Inventory (IRSI)
O4 = Interpersonal Communication Inventory (ICI)

model will be recommended to address the problem of below breakeven capacity during
the said period leads to other concerns, such as delayed flights, cancellations of flight,
and transfer of passengers to other airlines. After constructing a model, the group made a
program that would solve the model constructed. The program was made in order to
visualize the effectiveness of the solution to the program. The program will determine the
precise scheduling in order to minimize losses resulting from below break-even capacity
 flights. Lastly, before implementing the solution, the group presented the advantages and
disadvantages of the said model by constructing cost-benefit analysis.

A study called “Financing Higher Education” used action research design, as presented below:

The study employed action research design so as to enable the Commission on Higher
Education to formulate policies and standards to foster quality in higher education related
to:
1. Subsidy to private Higher Education Institutions;
2. Budget allocation scheme for public Higher Education Institutions;
3. Institutional cost;
4. Prioritization of resource allocation to various programs of public Higher Education
Institutions;
5. Cost of program by mode of delivery–formal or alternative;
6. Rationalization of tuition and other school fees, and;
7. Allocation of scholarship grants and slots for student and faculty.
The rationalization of these policies will make the generation and allocation of
resources more efficient and responsive to the needs of the Filipinos and the country.

4.2 Sampling

Sampling is the process of choosing adequate and representative elements from the
population. By studying the sample, the researcher is able to draw insights and conclusions for
the entire population.
In data gathering, sampling has many advantages over census or total enumeration. Aside
from being able to generalize the findings for the entire population, the researcher also able to
save in terms of time, effort, and cost. Sampling makes the scope of the study manageable
because of the small number of respondents to be covered, and increase the likelihood of
obtaining more reliable and accurate result.
For the reliability, samples, however, must meet the criteria of adequacy and
representativeness. Adequacy refers to the sample size (n), while representativeness pertains to
the possession of the characteristics of the sample as specified characteristics of the population.
The adequate number of elements to be taken as samples is based on the desired confidence level
(alpha:α) and room for error (e) in selecting the correct sample. In the academe, the most
common confidence levels employed in thesis and dissertation sample size computations are .01,
.05, and .10. The higher the confidence level desired, the bigger the sample size should be.
Statistics books contain different formulae in determining the sample size. However based on
actual experience, this formula which came from the Philippine Social Science Council Survey
Series (Publication Number 2) comes very handy:

n=

where:
1. at 95% confidence level Z α/2 = 1.96
at 90% confidence level Z α/2 = 1.65
 at 99% confidence level Z α/2 = 2.58
2. Assume proportion p=0.5
(as preliminary estimate)
3. Sampling error E normally assumes the
values: 0.025; 0.05, 0.10

The foregoing formula is often used for sample size determination in research involving
proportions (e.g., proportion of respondents willing to buy the products at a specific price,
proportion of satisfied employees.) As Vasquez(1997) pointed out, the formula is applicable to a
“finite population (or to one where sampling is done without replacement, as in simple random
sampling which is binomial or dichotomous in nature.” Since most business research studies
involve proportions, the said formula is often used.
Sample sizes for different population size were computed based on the foregoing formula to
come up with the table in Appendix A which could help the researcher in determining the sample
size (n) requirement of a given study, by simply referring to this table.
As a process, sampling involves these steps:

Sampling designs are commonly classified into probability and nonprobability sampling.
Probability sampling gives each element of the population a known and equal chance to be
included as a sample, while nonprobability sampling does not provide this predetermined chance.
Probability is used when inferences about the population are required, as in thesis, dissertation or
other academic researches. Nonprobability sampling is usually adopted when immediate
feedback is needed, as in marketing research studies, such as product launching. Specific
sampling techniques or procedures under each design are available to researchers (see figure 12).
Choice of technique is based on study objectives and hypothesis, the extent to which
generalizations are to be made, and cost and time limitations.

4.2.1 Probability Sampling

 Random Sampling affords each element in the population an equal and known
chance of being chosen as a respondent. The selection of the respondent can be done with the use
of the random table (refer to Appendix B), or a software program, by calculator, or through
lottery. The complete list of the population frame, e.g., list of registered small and medium
enterprises in the National Capital Region, list of the Top Corporations in the Philippines, list of
all residents in a particular village, and list of all regular rank-and-file employees on a firm, must
be available for use in simple random sampling.
Here are examples how random sampling is used.

EXAMPLE 1
This study aims to assess the effect of the currency crisis on the financial performance
of top corporations in the Philippines. The study made use of random sampling to give
each of the corporations on the Top 100 list for 1993 an equal chance of being included
in the sample. Based on appendix A, a population (N) equal to 100 necessities n = 79
samples for a 95% confidence level and a margin of error of 0.05.
The following steps are to be followed in the selection of the sample:
1. Use the table of Random Numbers (Refer to Appendix B) in selecting 79 samples.
A row and three columns from the table of random numbers will be chosen as
starting point At random, row 1 and columns 10 to 12 are chosen.
2. Using Appendix B, go down columns 10 to 12 from row 1, selecting the first 9
distinct numbers between 1 and 100.
3. Continue to the next rows and columns (13 to 15, 16 to 18 and so on) until all 79
distinct random numbers are chosen.
 4. These 79 random numbers represent the rank of companies chosen based on the
Top 100 list (Refer to Appendix C) chosen to be included in the sample.
For purposes of illustration, below are the first 10 numbers and companies chosen at
random.

RANDOM NUMBERS/
RANK OF COMPANY COMPANY
CHOSEN

49 Metro Pacific Corporation

25 United Laboratories Inc.
35 San Miguel Foods, Inc.
20 Landbank of the Philippines
24 General Milling Corporation
21 Metropolitan Bank and Trust Co.
62 Marsman and Company, Inc.
91 Philippine Post Authority
80 Nissan Motors Philippines, Inc.
26 Development Bank of the Phils.

*Based on Appendix C

EXAMPLE 2
This research was undertaken to determine how company employees react to news of
possible reorganization to adapt to market environment changes. Simple random
sampling was used to be able to select samples based on an alphabetical listing of the 350
employees. A sample size of 183 was derived through the use of the following formula:

n= = = 183

Samples were selected according to the following steps:

1. The alphabetical listing was numbered from 1 to 350.

2. To obtain 183 random numbers, the random function of the calculator was used.
3. Random samples were selected based on the random numbers obtained which
represent the numbers obtained which represent the number of the employee n the
alphabetical list.

For illustration purposes, the following are the first 10 significant random numbers that
came up through the use of the calculator as well as employees chosen as respondents
based on the list in Appendix D.
 Systematic Sampling allows the elements of the population to be selected as respondent
by using a constant number (k) derived from dividing the total population (N) by the
computed sample size (n). The formula is k = . Similar to random sampling, the
availability of the list of the population frame is necessary to apply this technique.

EXAMPLE
This study employed systematic sampling. The said method was chosen for two
reasons: (1) the population under study that includes all households in Sunrise Village is
described as heterogeneous in relation to the area under study, which aims at assessing
the feasibility of establishing a new meatshop in the area; and (2) an available listing of
homeowners, arranged according to block, made systematic sampling easy to execute.
To implement the systematic sampling method, two items need to be derived: (1) the
sampling interval (k) and (2) the random start (value between 1 to k). There are 1,600
houses in the population (N), therefore we obtain a sample size of 310 (based on
Appendix A). The sampling interval (k) was computed as:

k = N/n 1600/310 = 5 (rounded off)

Selection of sample was then employed according to the following steps:

1. Choose a random start. A random start of 3 was selected through lottery.
2. Use the listing of homeowners to number the households. The third homeowner on
the list was numbered as “1.” All houses were then numbered sequentially from 1 to
1,600.
3. To select the samples, the sampling interval of 5 was used. Houses with the
following numbers were chosen: 1, 5, 10, 15, and so on until 310 samples were
selected.
Note: Since the sequence in which items are ordered is important in systematic
sampling, replacement techniques were designed by sampling specialists. If one of the
selected samples (e.g., House Number 10 or House Number 25) is unavailable, the
following replacement strategies were suggested by Bojma and Atkinson (1995):

a) replacement in the first instance is the house before the house selected (House
Number 9) and second instance is the house after the house selected (House 26) and
so on.
b) replacement through flipping a coin (head-house before; tails-house after).

 Stratified Sampling helps in the selection of respondents where there are already
identifiable subgroups or strata of elements within the population which are of interest to the
researcher. A good example is when the research design is basically comparative in nature,
where the main objective is determining similarities or differences. Once the stratification
was made, the final respondents can then be selected either by random or systematic
sampling using proportionate or disproportionate number of elements. Stratifying the
population based on variables such as gender, age, income level, size of company, nature of
businesses, ownership of firm, establishes the homogeneity within each subgroup so that
clear differences between groups are determined.

EXAMPLE 1 (Proportionate Allocation)

This study intends to survey the effects on employee satisfaction of cost-cutting
measures implemented by the company. The study also aims to determine if there are
differences in responses among rank-and-file employees, middle managers, and senior
managers. To attain this objective, the study used stratified random sampling with
proportional allocation. Each level (Rank and File, Middle Management or Senior
Management) represents a stratum. The table below shows the population for each
stratum, proportional allocations, and sample size needed.

Sample Size Allocation

Among Different Position Levels of Employees

Level Subpopulation Percentage Sample

Size () Distribution Size
()

Rank and File 150 75.00% 99

Middle Management 30 15.00% 20
Senior Management 20 10.00% 13
TOTAL 200 100.00% 132

Samples from each stratum were selected through random sampling, making use of the
Table of Random Numbers and the listing of employees grouped by level/position. A
major consideration in the choice of stratified random sampling the characteristic of the
population itself of having homogeneous groups for each stratum or subpopulation as
well as the sampling method’s applicability to the nature of the problem that the study
would want to focus on.

EXAMPLE 2 (Proportionate Allocation)

The study, in line with its objective of assessing the effects of the economic crisis on
the sales performance of insurance companies, used a stratified sampling method. Since
insurance companies are either life or non-life, this distinction might possibly pose
differences in the results of the study. The Philippine Financial Almanac was used as
reference for the sampling frame of insurance corporations. Proportional allocation and
sample size were obtained as shown in the following table.

Sample Size Allocation for Life

and Non-Life Insurance Companies

Classification Subpopulation Percentage Sample

Size () Size
()

Life Insurance 31 23.85% 23

Non-Life Insurance 99 76.15% 74
TOTAL 130 100.00% 97

Random Sampling was used in the selection of samples for each stratum.

EXAMPLE 3 (Disproportionate Allocation)

 In a similar study as that in Example 1 but for a different company, using a
proportionate sampling method would result in the following:

Sample Size Allocation

Among Different Position Levels of Employees

Level Subpopulation Percentage Sample

Size () Distribution Size
()

Rank and File 1455 97.00% 297

Middle Management 30 2.00% 6
Senior Management 15 1.00% 3
TOTAL 130 100.00% 306

Analysis of the above table indicates that on the level of senior and middle
management, the 9 respondents may not truly represent their group. The sample size is
therefore too small. In this case, disproportionate allocation may be more appropriate.
Sound judgment may point to the following recomputations of sample size per stratum.

Sample Size Allocations

Among Different Position Levels of Employees

Level Subpopulation Size () Sample Size

()

Rank and File 1455 283

Middle Management 30 15
Senior Management 15 8
TOTAL 1500 306

 Cluster Sampling involves the grouping od division of the elements of the population
into heterogeneous groups. Then some of these groups are randomly selected and all the
elements of the cluster are studied. It should be noted that each cluster sample is composed
of respondents with different perspectives and interests. This way of selecting is deemed lees
costly. Some normally occurring clusters in business research are clusters of managers,
customers, suppliers, stock brokers, investors, accountants, and advertisers. Since the
elements do not have much differences among them, it appears that the results show less
precision.
Here is an example of cluster sampling.

The company has implemented Total Quality Management through the creation of 98
quality circles made up of 10 people each representing employees from different sectors
in the Production Department. This made sure that each quality circle is well-represented
in terms of issues that can be raised during quality-circles meetings. The Production
Manager would want to assess the effectiveness of Total Quality Management along
different sections on his department. To be able to do this, the study made use of cluster
sampling for its sampling method. The choice of this particular method was based on the
characteristic of each cluster (quality circle), which is heterogeneous. Each cluster of
quality circle represents all employees from different sections.
Based on a population of 975 employees, a sample size of 276 was obtained. This,
therefore, resulted in the selection of 28 quality circles. A list was made of all 98 quality
circles, and 28 quality circles were interviewed and considered in the study.

 Area Sampling pertains to the grouping of the population into geographical divisions
before selecting the respondents. This sampling can be done if there exist a clear delineation
of communities where the respondents can be found. Common sampling areas are villages,
subdivisions, cities, and municipalities.
Here is an illustration of how area sampling is undertaken.

The study made use of area sampling for its sampling method. The study aims to
determine extent of usage of Direct Distance Dialing service among 10,000 households in
375 blocks of houses in interconnected villages in Laguna. The said population is known
to be heterogeneous based on previous studies made. Area sampling was used by (1)
randomly selecting the blocks to be included in the study and (2) randomly selecting the
elements from each of the blocks.
From 375 blocks, 190 blocks (2,850 households) were chosen in the sample (based on
appendix A). From 190 blocks, again random sampling was used to get a sample size of
370 (based on appendix A, N = 10,000).
To get the number of samples to be selected from each of the blocks, the number of
houses per block was taken into consideration. The number of samples from each block is
proportional to block size (e.g. Block 23 was chosen among 190 blocks to be sampled
and it has a total number of 25 households which is 0.88% of 370 is 3, three households
were randomly chosen for block 23. The same pattern of computation was followed for
other blocks.).

 Double Sampling means getting a smaller sample from the initial large sample. This
design is sometimes called sample within a sample. Double sampling is usually done when
the researcher intends to gather more in-depth and focused data on the topic of investigation.
The initial larger sample provides preliminary information which helps in determining the
second sample set to be drawn from the same sample group.
Here are some examples.

EXAMPLE 1
In the absence of current information on the household size of those residing in Sun
Valley Subdivision, the proponents of this study decided to employ double sampling
 method. Household size is very crucial to the study whose objective was to compare the
differences in effects of the economic crisis on household expenditure patterns of
households of size 5 members and less and that of 6 members and more. Simple random
sampling was done first. Based on the population of 500 households, the sample size
obtained was 217. Samples were selected through the use of the Table of Random
Numbers. From this initial example, a survey was made on 217 households selected by
random sampling. Data on household size was obtained along with other data needed for
the study.
After obtaining the said data, estimates were made regarding proportion of households
classified by size as shown in columns 1 to 3 of Table 1.
Stratification was employed, while proportional allocations and sample size were
computed for the second phase of sampling as shown in column 4 of Table 1.

Table 1
Estimated Sample Size for Sun Valley Subdivision,
By Household Size

Classification Subpopulation Size Percentage Sample

() Size
()

Households (5 126 58.06% 81

members or less)
Households (6 91 41.94% 59
members or more)
TOTAL 217 100.00% 139

Based on proportional allocation and sample size, 139 samples were selected from the
initial 217 samples to make the analysis more relevant based on the objectives of the
study. There were 81 households with 5 members or less and 58 households with 6
members or more considered for the study.

EXAMPLE 2
The researcher wanted to know the assessment of retailers regarding the liberalization
of the retail industry. Random sampling was initially used among retailers in Metro
Manila. During the first phase of the study, awareness regarding the issue was first
determined as well as the general profile of retailers through a self-administered
questionnaire. Through the results of the initial phase, the researcher was able to identify
the retailers who had high awareness of the issue. There was a total of 82 retailers who
had high awareness in the issue. In the second phase of sampling therefore, sampling,
was made among these 82 retailers. Due to the cost constraints, the proponents were only
able to conduct interviews among 30 samples. Random sampling was conducted to select
the 30 samples for the second phase of the study.

 Multi-Stage Sampling is cluster sampling done in several stages. This type is usually
used in nationwide surveys, which each region, province, city, municipality, or barangay is
subdivided and selected, before the actual respondents are chosen.

EXAMPLE 1
 A nationwide survey will be undertaken for a study to assess and measure the extent of
unemployment among Filipino households. To be able to achieve this, a multi-stage
sampling is appropriate for the selection of sampling units (for this study, it is the
provinces/municipalities within the region chosen). Secondary sampling units are the
towns/municipalities. Tertiary sampling units are the barangays within the
towns/municipalities chosen, while the ultimate sampling units are the households.
Below is the table showing the population sizes of every region.

Regions/ Areas # of Provinces/ Number of %

Domains Municipalities Households

NCR
CAR Metro Manila 10 1,765,644 14.21
I Northern Luzon 6 241,204 1.94
II Ilocos 4 706, 263 5.68
III Cagayan Valley 5 523,098 4.21
IV Central Luzon 11 1,274,646 10.26
V Southern Tagalog 15 1,731,396 13.93
VI Bicol 6 877,895 7.07
VII Western Visayas 7 1,133,399 9.12
VIII Central Visayas 6 952,353 7.66
IX Eastern Visayas 5 693,679 5.58
X Western Mindanao 4 508,768 4.09
XI Northern Mindanao 9 734,195 5.91
XII Southern Mindanao 7 887,145 7.14
ARMM Central Mindanao 6 395,243 3.18
Autonomous Region of 4 330,016 2.66
TOTAL Mindanao
12,424,928 100.0
0
*Based on 1994 census

Based on this table N = 12,424,928. Computations based on the formula that we have
been using lead to a sample size of 384. After getting the population and sample size, we
obtain the sampling fraction, which is:

f= = =

This sampling fraction would be used in the selection of households. This indicates that
assuming equal overall probabilities of selection, the rate is 1: 32,357.

In the first stage of selecting regions are divided into domains.

Based on the first table shown, a possible grouping into domains can be based on
proximity and sample sizes so that we have the following domains and first-stage
selection probabilities.

Regions/ Areas Number of Subtotal 1st Stage Selection Selection

Domains Households s Prob. Range
Domain 1,765,64
1 Metro Manila 1,765,6444 4 1:1 Certainty
NCR
Domain Northern Luzon 241,204 1,470,56 1:6 1
2 Ilocos 706,263 5 1:2 2,3,4,
CAR Cagayan Valley 523,098 1:3 5,6
I
II Central Luzon 1,274,646 1:1 Certainty
Domain 1,274,64
3 Southern Tagalog 1,731,396 6 1:1 Certainty
III Bicol 877,895 -
Domain 2,609,03
4 Western Visayas 1,133,399 2 1:1 Certainty
IV
V Central Visayas 952,353 1:2 1
Domain Eastern Visayas 693,679 1,133,39 1:2 2
5 9
VI Western Mindanao 508,768 1:2 1,2,3,4,5
Domain Central Mindanao 395,243 1,646,03 1:3 6,7,8
6 Autonomous Region of 330,016 2 1:4 9
VII Mindanao
VIII
Domain Northern Mindanao 734,195 1,234,02 1:2 1
7 Southern Mindanao 887,145 7 1:2 2
IX 12,754,944
XII
ARMM

Domain 1,621,34
8 0
X
XI
TOTAL

First-stage selection probabilities were obtained based on sample sizes per region per domain. It
represents the probability of getting the region as representative of the said domain. Selection range was
used in the random sampling method (lottery). To illustrate for Domain 2:

CAR’s sampling fraction was obtained as

f= = =

Ilocos :

f= = =

Cagayan Valley :

f= = =
In the lottery, the selection range was assigned based on the first-stage selection probabilities. For
example, for Domain 2, the lottery resulted in selecting number 4. This would mean that Ilocos is chosen
as the region to represent Domain 2.

Overall, there is a total of 8 regions representing the 8 domains included in the sample. NCR, Western
Visayas, and Central Luzon are certain to be included in the sample because of their sample sizes.

Later Stages
In the succeeding stages, the basic point is that the product of the first stage and subsequent stage
selection probabilities (second, third, and so on) should result in the same overall fraction, which is
1:32,357 (taken from the computation shown on page 73).

To illustrate, if Ilocos was chosen, then selection of provinces, municipilaties, barangays, and households
in the subsequent stages in the Ilocos region would have the following selection probabilities.

REGION First Stage Selection Succeeding = Overall

Probability x Stage(s)
Example:
Ilocos 1 x 1 = 1
_______ ________ _______
2 161,785 32,357

In the next stages, therefore, each region selected should be divided into provinces or municipalities (e.g.,
NCR) in the second stage, barangays in the fourth stage, and finally households in the fifth stage. The
same step done in the initial stage can be used for the subsequent stages. Groupings or stratum (similar to
what was done in selecting domains) can also be made based on certain analytical considerations of the
researchers. In the implementation of the selection process, it is best to consult a sampling expert.

Note: Unequal probabilities of selection may be used. Regarding multistage sampling, the book The
Sample Survey: Theory and Practice by Lininger and Warwick (1975) is a good reference.

EXAMPLE 2
In addition to the first example, national surveys made by National Statistics Office (NSO) have
introductory discussions on publications. NSO uses multistage cluster sampling in its nationwide surveys.
And this is how the multistage cluster sampling was explained:
The sampling design of the 1994 Family Income and Expenditure Survey (FIES) adopted that of the
Integrated Households of the NSO, which used a stratified two-stage cluster sampling.

Domains
The urban and rural areas of each province are the principal domains for the survey. In addition, areas
with 150,000 or more populations as of 1990 are also domains in the survey with urban and rural
dimensions. These include the four cities (Manila, Quezon City, Pasay, and Caloocan) and five
municipalities (Valenzuela, Paranaque, Pasig, Marikina, and Makati) of Metro Manila; other key cities
such as Angeles, Baguio, Cabanatuan, Olongapo, Batangas, Lipa, Lucena, San Pablo, Bacolod, Iloilo,
Cebu, Mandaue, Zamboanga, Butuan, Cagayan de Oro, Davao, General Santos, and Iligan; and key
municipalities such as San Fernando, Pampanga, and Tarlac.
The remaining municipalities of Metro Mnila (Taguig, Mandaluyong, San Juan, Pateros, Las Pinas, and
Muntinlupa) were treated as separate domain known as the other Metro. It should also be mentioned that
in the case of Makati, six exclusive villages were identified and samples were selected using a different
scheme. These villages are the Forbes Park, Bel-Air, Dasmarinas, San Lorenzo, Urdaneta, and
Magallanes.
Because of the creation of the Autonomous Region of Muslim Mindanao (ARMM), Marawi City and
Cotobato City were likewise treated as domain.

Sampling Units
The primary sampling units (PSUs) under the sample design are barangays, which were classified as
either urban or rural. Households within each sample barangay comprised the secondary sampling units
(SSUs). Within each domain, a pre-determined number of PSUs (based upon approximately “proportional
to size” sample allocation scheme) was selected at the first stage of sample selection procedure with
probability proportional to size (PPS). Prior to selection, PSUs in each domain were ordered by
population size, creating a secondary (i.e., implicitly) stratification intended to improve the efficiency of
the sample design. Population counts from the 1990 Census of Population and Housing (CPH) were used
as measures of size during first stage sample selection.
Within each PSU selected at the first stage, a predetermined number of households (SSUs) was selected at
the second stage using a systematic random selection procedure with equal procedure with equal
probability. The number of households chosen from each PSU took into account the probability of
selecting the PSU at the first stage such that of each household within the domain had an equal probability
of selection for the survey, that is, the sample was self-weighting within domains.

= x x
where:

= number of sample households allocated to each PSU

= raising factor for the kth domain (urban/rural area of province, city, or special area)
= number of households listed in the stratum
= 1990 population total for the stratum
= 1990 population of the stratum
= number of sample PSUs in the stratum

Sampling Fraction
In general, the sample design and results in self-weighting samples within domains, with a uniform
sampling fraction of 1:400 for urban and 1:600 for rural areas. However, special areas are assigned
different sampling fractions so as to attain adequate samples for each. Special areas refer to urban and
rural areas of province or large city which are small relative to their counterpart. These special areas and
their sampling fractions are as follows:

AREA Sampling AREA Sampling

(Urban) Fraction (Rural) Fraction

Batanes 1/30 Butuan City 1/300

Siquijor 1/60 Aurora 1/300
Ifugao 1/90 Southern Leyte 1/360
Mt. Province 1/90 Iligan City 1/360
Marinduque 1/120 Batanes 1/60
Quirino 1/120 Cabanatuan City 1/90
Camiguin 1/150 Cagayan de Oro City 1/120
Catanduanes 1/180 Camiguin 1/240
Kalinga-Apayao 1/210 Batangas City 1/300
Abra 1/240 Lipa City 1/300
Romblon 1/240 San Pablo City 1/480
Basilan 1/240 Aurora 1/570
Tawi-Tawi 1/300
4.2.2 Non-Probability Sampling

 Quota Sampling is selecting the predetermined required number from the population
regardless of how they are chosen. This design is usually done in opinion or poll surveys. An
example is shown below.

A study wants to get feedback among citizens in a particular municipality regarding an

advocacy advertising recently conducted by a company in different residential areas in
the said municipality. One of the objectives of the study is to get different opinions from
those coming from the lower-income, upper-income, and high-end income households of
the municipality. The population has a total of 1,500 households.
Quota sampling was used through the following steps:

1) Stratification of the population. Based on previous studies made, the following are
the percentage distribution among different income levels in the municipality: low
income (60%), middle income (30%), and high income (10%).

There are also known areas (barangays and villages) where household of particular
income segments can be found.

2) Predetermining of quota. Based on budget considerations, only 100 respondents

could be interviewed–60 from low-income, 30 from middle-income, and 10 from high-
income households.

3) Selection of samples. Interviews were then conducted among: (a) the first60
respondents from low-income households available for interview and found in the low-
income barangays; (b) the first 30 available respondents from middle-income households
from the middle income villages; and (c) the first 10 available respondents from high-
income households found in high income villages.

 Judgment Sampling involves the selection of respondents considered to be in the best

position, or most knowledgeable, to give the needed information. Some data about the company
are known only to key informants, such as the controller, managers, or owners of the firm.
Here is an example.

A study wanted to conduct an assessment and analysis of the applicability and use of
environmental cost accounting in Philippine industrial companies. Since the topic under
 study was very specialized, the researcher opted to use judgment sampling. Judgment
sampling was implemented to select the companies to study, and to select the respondents
from each company. First, industrial companies likely to implement, or are
implementing, environmental cost accounting were selected (e.g., mining, cement,
shipping companies). Second, from these companies, either the Accounting Manager or
the Chief Accountant was selected to become the respondents since they have the
experience and extensive knowledge regarding the matter.

• Convenience Sampling allows the researcher to gather data from respondents who are
conveniently available to provide the necessary information. This design is used to get
information fast.

To illustrate:

EXAMPLE 1
The study wanted to assess customer satisfaction in the Mindanao area, particularly to
owners of the brand new car model launched by the company. There are 456 new car
owners in the area. The selection of the sample was made through convenience sampling
by interviewing all new car owners who dropped by the service center of the company for
maintenance and check-up. The said interviews were done for five successive working
days.

EXAMPLE 2
To assess the effectiveness of a training program a study employed convenience
sampling to get immediate feedback regarding the matter. In a general meeting of the
company, researchers employed convenience sampling by interviewing those who
attended the training program among those who attended the meeting.

• Accidental Sampling implies that the information is collected from respondents who, by
chance or circumstance, are met by the researcher in the process of data gathering.
Here is an example of a situation that used accidental sampling.

The study concerns the perceptions of foreign buyers on Philippine export products.
Accidental sampling was employed by interviewing foreign exporters that the researcher
met by chance during the trade fairs held in the Philippines for the past year.

• Snowball Sampling means choosing initial samples. These samples will then refer
other respondents from whom the same information may be obtained.
Below is an example of snowball sampling.

This research is about the “Perception of Foreign Investors on the Future of the
Philippine Economy”. To do this, the study employed snowball sampling for its method
of sample selection. The researcher initially interviewed a foreign investor with whom he
was able to set an appointment. At the end of the interview the researcher requested to be
referred to respondents whom the first interviewee deems appropriate to answer the same
questions asked of him. The same method was applied for succeeding interviews
conducted.
• Purposive Sampling involves the selection of key informants based on a predetermined set of
criteria. These are people considered to be the most appropriate source of data in terms of the
objectives of the study. To illustrate:

Purposive sampling was used to determine the profile of the households with female
family heads (defined as those who provide financial support to the family, with the
major bulk of household income coming from them). The following steps were
undertaken:
1) Setting of a criteria for the selection of respondents. The respondent is a female
family head of households with household size of 5 or more. She earns equal to or more
than P 120,000.00 a year.
2) Selection of sample based on the preceding criteria.

4.3 Methods of Data Collection

The researcher has two major options in collecting data: (1) primary data collection
(e.g., survey, observation) and (2) secondary data collection.

4.3.1 Primary Data Collection

This method involves the gathering of data directly from the respondents through the use of
different techniques. The common techniques applied in business research are the survey (either
using questionnaire or interview), observation, tests and scales, and experimentation.

4.3.1.1 Survey. The use of survey as a method of data collection has gained popularity over
the years because of the great deal of valuable information on what people think about certain
issues, products, policies, or events. The survey method, if done correctly, is a relatively easy
and quick way of finding out people’s response to the research problem being investigated.
The common survey techniques used in data collection are personal interview, telephone
interview, self-administered questionnaire, mailed questionnaire, and e-mail/Internet. Table 12
presents a brief comparison of each technique. The descriptions on personal interview, telephone
interview, and questionnaire are based on Clover and Balsley (1979).

(P.S. Table 12 will be on a different file due to its orientation..ty )

Interviews or questionnaires may be used when using the survey as data-gathering procedure.
In both cases, a form or a research instrument is used to solicit the responses from individuals.
Table 13 summarizes the comparison.

Table 13. Comparisons of Survey Form

TYPE DESCRIPTION

QUESTIONNAIRE  Self-administered. Respondents reads and

(Example is shown in Appendix E) answers the questions
 Either mailed or non-mailed
 Contains a written introduction about the
study and a detailed instruction
  Predominantly contains fixed alternative
questions predetermined by the researcher.
INTERVIEW SCHEDULE  Questions are personally asked by the
(Example is shown in appendix F) interviewer
 Sequence of asking questions is rigid
 Predominantly contains open-ended or
unstructured items or list of topics
 Questions are directly asked by the
researcher but allows the respondent to
answer the questions in whatever manner
one prefers
INTERVIEW GUIDE  Predominantly contains open-ended or
(Example is shown in Appendix G) unstructured items or list of topics to ask
 Questions are directly asked by the
researcher but allows the respondent to
answer the questions in whatever manner
one prefers
 Flexible

Since the questionnaire is widely used in surveys, the following guidelines can help the
researcher in designing a good one.
 Content Criteria

 Purpose. The items to be included in the form should adequately answer the objectives and
hypothesis of the study. When the respondents’ buying preferences are being measured, the
questions listed should tap the components and elements of the variable. Therefore, the items to
be appropriately included, according to Sedlack and Stanley (1992), must have face validity,
unidimensionality, response variation, and item format. Question then must be related to the
variable being investigated, must assess only one dimension, must generate an array of answers,
and must be stated either in declarative or question form.

 Language. The words to be used in the questions should be familiar and nearest to the
respondent’s level of comprehension. The use of jargons and idioms normally used by the
researcher but not by the respondent should be avoided. Words like reengineering, harmonization
of accounts, pareto principle, and horizontal and vertical integration should not be used in a
survey on blue-collar rank and file.

 Type of Questions. There are two basic types of questions which can be utilized to solicit
responses. The fixed alternative (structured) and open-ended (unstructured). The fixed
alternative question provides an array of response categories from which the respondent will
choose or base one’s response (s). For instance, the respondent is asked to rate one’s level of job
satisfaction using a five-point Likert Scale, which indicates:

5 4 3 2 1

Very Satisfied Neutral Dissatisfied Very

Satisfied Dissatisfied

In another instance, the respondent is asked to encircle one’s frequency of buying blue jeans
from a department store using these choices:
1 Every month
2 Every 2 months
3 Every 3 months
4 Every 4 months
5 Every 6 months
6 Once a year

This type of question is advantageous to use when the set of responses are few, finite, and
clearly delineated. This type enables the respondent to easily recall and make a fast decision
regarding the answer based on choices. Therefore, this type is less time-consuming and easier to
tabulate, especially if the set of alternatives is pre-coded. On the other hand, the fixed alternative
question tends to limit the answer of the respondent to the choices provided in the set. The other
limitation of this type of question is that it cannot be used to probe complex emotions. It cannot
also be used for questions with no finite set of answers.
The open-ended question, however, gives the respondent the opportunity to respond to the
question in the manner one prefers. For instance, the respondents may be asked to cite significant
factors that made their company successful. The respondents may be asked about what they look
for in patronizing a restaurant. Or respondents may be asked about their feelings regarding
management style of their firm. This type of question should be used when the possible answers
are varied, numerous, and can possibly overlap. This type enables the respondent to give answer
based on one’s own perspective, allowing one to probe deeper into one’s emotions. However,
this type of question is time-consuming and requires the development of categories before data
can be tabulated.

 Order. The questions should be arranged in a manner which facilitates the respondent’s
replies, avoiding biased responses to the item that follows. Hence, items can be sequenced from
general to specific, easy to difficult, simple to complex, or less serious to more serious
conditions. There are also instances when a filter item is necessary to minimize respondent
difficulties in answering question of which they have no sufficient knowledge. More importantly,
the filter item seeks to segregate respondents who are required to answer a series of question
from those who do not find such items relevant ( Sedlack & Stanley, 1992). For instance, the
researcher wanted to know if a sample of consumers would buy a CD rack. If respondents do not
own CDs, it would be pointless to ask them if they will buy one for themselves. Here the filter
question “Do you own CDs?” will identify those consumers who own CDs from those who do
not. In the sequence of questions, it is also advisable not to put positively and negatively-worded
questions measuring the same dimension of a variable in consecutive order (Sekaran, 1992). To
illustrate two questions, such as the following, are put, one right after the other:

1. My boss allows me to be innovative in my job.

2. My boss does not allow me to be innovative in my job.

If there is really a need to cross-validate answer consistency, these items should be placed in
different parts of the instrument. Moreover, the researcher sometimes strongly believes that all
the items are equally important and no question will render biased the respondent to any
subsequent item. In this case, the item should be randomly placed using a table of random
numbers ( Sedlack & Stanley, 1992). Lastly, in determining the arrangement of the questions, it
is also recommended to put together or place in the same block, similar or related items
measuring the same variable or set of variables. Within the same block, the other criteria should
be observed by the researcher. Refer to Appendix F for examples.

 Measurement. After the items have been developed, it is important to determine the type of
response categories that will be used for fixed alternative questions. The most commonly
adopted response category is the Likert Scale (e.g., 5 – Strongly Agree, 4 – Agree, 3 – Don’t
Know, 2 – Disagree, 1 – Strongly Disagree). Once the response categories have been developed,
it is helpful to code the responses. This process will reduce the time of the respondents in
selecting the answer to establish the validity and reliability of the instrument. Validity pertains to
the ability of the questionnaire to accurately measure what it intends to measure based on the
operational framework. This ability can be established by face or by content validity and
reliability of the instruments. For example, company profit can be measured by deducting total
expenses from total revenue or by computing the ratio of ROI. Reliability pertains to the level of
consistency that a measurement possesses. It means that when a measure is applied repeatedly on
the same set of respondents, similar results will be obtained. One of the most popular tests of
reliability is the test-retest procedure through an experimental design. For instance, take two or
more similar groups of respondents. One is assigned as the experimental group which is given
instructions at two different times (test and retest). The other group is designated as the control
group which is given the instruction only during the retest period. The scores obtained are then
compared.

 Other Features

 Introduction. Since the questionnaire is self-administered, it must contain an introductory

part. This introduction should briefly state the purpose and significance of the study, the reason
for choosing the respondent, assurance of data confidentiality, and gratitude to the respondent for
his/her cooperation. The researcher’s official and e-mail address, telephone and fax numbers,
must be indicated to facilitate communication between respondent and researcher.
 Appearance. The external features of the questionnaire should motivate the respondent to
take a look at the instrument. Therefore, the form should be neat, organized, properly typed,
devoid of typographical errors, and adequately spaced. Enough space must be provided
responses to open-ended questions. Even the type of paper to be used (e.g., bond, mimeo,
newsprint) should be appropriate to the kind of respondents. In any case, recycled paper with
writings on one side should never be used. Instruction should also be properly to avoid
confusion. For instance, specify if the respondent has to check (), cross (X) or encircle (o) the
answer. Also, if the respondent has to rank several responses by using 1 for the highest and 10
for the lowest rank, such instruction should be very clear.
 Length. This feature pertains to both the question and the instrument. Short questions are
preferred to lengthy ones. Each item should be in simple sentences, too. Simplicity and brevity
will facilitate respondent’s understanding. If the items are also validated, only the relevant
questions will be included. Thus, unnecessarily long instrument can be avoided. People in
business are very busy. The shorter the questionnaire, the lesser time it needs to be answered. To
determine more accurately the length of time needed in actually answering the form, the pretest
can be very useful. The pretest also enables the researcher to detect problems with items, such as
leading questions, those which are ambiguous, poorly worded, difficult to recall, or answer, and
those that are not answered at all. The pretest result will help increase the reliability and validity
of questions when properly analyzed and revised. Therefore, pretest should be conducted with
small selected group of respondents who possesses the same characteristics as those who will be
chosen for the study. Respondents for the pretest will not be part of the actual study.
 Personal Data. It is standard practice that demographic information such as age, civil status,
gender, educational attainment, occupation, and income are obtained. Whether to ask for this
data at the start or at the end of the questionnaire is the researcher’s option. However, asking the
name of the respondent to be directly written on the form should only be done when absolutely
needed. It is better to secure their names and write these on a separate sheet, while the form is
pre-numbered to match the list of names for cross-checking purposes. There is also another view
that even demographic information should only be solicited if they are relevant to the purpose of
the investigation.

4.3.1.2 Observation. It is possible to collect primary data without directly posing the
questions to the respondents. Such method requires observing people by systematically looking
at behavioral phenomena with the use of one’s senses and or with the aid of a camera, one-way
mirror, and/or recording instrument in observing people.
Observation as a method is advantageous in studies when original data can be collected at the
time it is happening. This method captures the entire event as it unfolds in a natural or man-made
setting when participants ignore more information which they consider common. Some good
examples of situations where observation is beneficial are in researches on consumer reaction
with regards to the taste of a brand new pizza or fried chicken or a new film shown in a
neighboring theater.
Observation, when properly planned and implemented by trained personnel is very valuable
research method. Hence, the researcher can decide whether to use participant or non-participant
observation as an approach. Participant observation enables the researcher to assume a
legitimate social role within the group being investigated, like becoming a member of an
organization that analyzes leadership styles and conflicts. On the other hand, nonparticipant
observation enables the researcher to be detached and need not to become an integral part of the
group under study, like observing consumers in a supermarket when they buy a particular brand
of laundry soap or studying staff behavior during a departmental meeting. In any of these
approaches, the researcher has the choice to whether to make known to the group that a research
is being conducted on them or to conceal the purpose of the research altogether.
 However, observation as a method, is time consuming and costly. It requires the researcher to
be physically present in the time of data collection, if a mechanical devise is not used. It is also
possible at the time of data collection; the participant’s behavior may not be normal one just to
accommodate the researcher. Respondents may also be withhold or limit information if they do
not want to cooperate with the observer. This is where the process of “triangulation,” as
endorsed by Webb, et al. (1981) can be used to respond to this situation. This process means that
a combination of different procedures (e.g., interview, observation of nonverbal behavior, and
observation of verbal behavior) from the same group of respondents can strengthen confidence in
one’s results since the data from one procedure can be validated with the data of the other
procedure. For studies of long duration, one way of strengthening the reliability and validity of
the results is to disregard the data in the initial periods if they appear to be different from the rest
of the observed data.

4.3.1.3 Other Data Collection Methods. Aside from the previously discussed methods of
data collection, the researcher can also use the following:

 Delphi Method is a qualitative process of acquiring information on issues. It involves

forecasting or projecting trends or outcomes: e.g., prospects of the banking industry given the
liberalization law, forecasts on the future of the economy given the Asian financial crisis, and
future of the labor unions given the globalization trend. The Delphi method entails choosing a
panel of key informants who are considered experts on the topic under study, to be the
respondents.
 Projective Method is the use of standardized psychological tests, such as inkblots, sentence
completion, and thematic apperception (TAT), to probe deeper into the minds, behavior, and
attitudes of respondents. These data can be very useful in studies on consumer preferences and
workers’ motivation. To apply this method, the service of trained or professional
psychometricians in administering the test will be required.
 Unobtrusive Method is using other data sources aside from individuals. For instance, the
time record is a source of information on employee attendance. The number of various brands of
bottled water found in trash cans is a measure of bottled water patronage level. Another good
example is the wear and tear on magazine rental store which could be a good indication of their
popularity. Lastly, the number of finished product rejects in the production department discloses
the level of effectiveness of the production line employees.

4.3.2 Secondary Data Collection

In data collection, the secondary method may be used which involves the acquisition of data
from secondhand sources like statistical publications, census reports, financial statements, annual
reports, brochures, catalogues and other documents and reports. These data are generated or
produced or kept by government agencies, private corporations, business and trade organizations,
labor organizations, educational institutions, and commercial information service groups.

(Please insert Figure14)

 In acquiring the needed data from these sources, it is imperative to consider their credibility,
accuracy, objectivity, timeliness, cost, and willingness to provide information which can be used
for the purpose of the research.
Looking at the various methods of data collection, it can be noted that each has a particular
inherent limitation. Since the limitation of the method can be assured of the reliability and
validity of the results. While the multimethod technique is often costly and time consuming, the
significant outcome of the investigation far outweighs its costs.

4.4 Methods of Data Analysis

Data analysis involves the application of the appropriate statistical tools to generate results
which can be interpreted meaningfully to answer the research problem posed at the beginning of
the investigation.
The most common problem of a researcher at this stage of the research process is choosing
the most appropriate statistical tool for data analysis. Available books on statistics reveal that the
choice of a particular tool of analysis will depend on the type and level of data, hypothesis
statement, sampling design and the requirements and assumptions of the statistical tools. Figure
15 shows some considerations in data analysis, while Table 14 presents the summary of the
guidelines which help determine the appropriate data analysis tools that can be employed given
the varying conditions in the study.

(Please insert Figure15)

Table 14 presents statistical tools that can be used depending on the purpose of the study
(descriptive or inferential) and on the methodology used.
Methodology is based on the nature of data and the sampling plan. The following
characteristics of data are considered before analysis:

1. Type (numerical or categorical)

2. Level (nominal, ordinal interval, ratio)
3. Number of variables analyzed at one time (univariate if one variable is analyzed at one
time, or multivariate if two or more variables are analyzed at one time)

Data analysis also considers the sampling technique used, i.e., whether the samples were
randomly chosen (using probability sampling techniques) or not. The analysis should also note
whether the sample size is small (n<30) or large (n>30). “Sample groups” that are formed and
analyzed should also be considered data analysis. Sample groups refer to the grouping of
elements of the sample based on certain characteristics (e.g., gender, age groups, civil status).
Table 15. Statistical Software for Data Analysis

PROGRAM DESCRIPTION
A simple statistics package originating in
MINITAB™ statistical education and available across a wide
range of computer types. It tis probably the most
popular statistics package and is easy to learn
and fairly flexible.

(Source: http://www.minitab.com)
A statistics package originating in the social
sciences and available across wide range of
computer types. It is particularly suitable for the
analysis of the designed experiments and
regression problems, but is not particularly easy
to learn.

Statistical Capabilities:
(Source: http://www.spss.com)

Frequency tables, measure of central tendency

Statistical Package for Social Sciences and dispersion, measures of skewness and
(SPSS) kurtosis, percentiles, histograms and charts,
computation of z or standard scores, boxplots,
Kolmogorov-Smirnov tests, crosstabs, nominal
data measures, measures of linear association
tests of independence, Mc Nemar’s tests, test of
linearity, t tests one-way anova, simple factorial
anova, correlation analysis, partial correlations,
linear regression analysis, Durbin-Watson tests,
chi-square tests, runs tests, Mnn Whitney test,
Kruskall Wallis tests, Wilcoxin test, Cochran’s
Q test, Friedman, Kendall’s tests, factor analysis
with rotation methods, cluster analysis,
discriminant analysis, multivariate analysis of
variance, loglinear models, multiple regression,
logit models, Box Jenkins

A statistics package originating in

GENSTAT™ agricultural science and available across a
moderate range of computer types. It is
particularly suitable for the analysis of designed
experiments and regression problems, but is not
particularly easy to learn
PROGRAM DESCRIPTION
A collection of statistics programs originating
BMDP™ in the bio-medical sciences and available across a
moderate range of computer types. It is widely
used in medicine and is very useful if your
problem fits one of the programs, but it is not
very flexible.

A very large and sophisticated data handling

SAS™ package, which is most suitable for people spend
a lot of their time manipulating and analyzing
data. It is available on a number of different
computer types. It takes a fair amount of effort to
use SAS and is therefore not very suitable for
casual use.

Statistical Capabilities
(Source: SAS Institute, Inc.)

Analysis of variance (one-way, n-way, repeated

measures), analysis of covariance, means
comparison, t tests, hierarchical clustering,
disjoint clustering, experimental design, runs
tests, pareto charts, Shewart charts, real time data
collection, chi-square tests, odds ratios, conjoint
analysis, corresepondence analysis, cross-
tabulatios, log-linear models, logistic regression,
multidimensional scaling, multidimensional
preference mapping, box-plots, cube plots,
histograms, scatterplots, canonical discriminant
analysis, Mahalanobis distance, stepwise
discriminant analysis, discriminant analysis,
canonical correlations, correlation analysis, factor
analysis, partial correlations, principal
components, canonical response surface analysis,
logistic regression, Poisson regression,
polynomial regression, univariate regression,
weighted least squares regression, multivariate
regression, ridge regression, distribute tests,
measures of central tendency, quantiles,
deviations, variance

A statistics package originating in psychology

SYSTAT™ and available on PCs and VAXs only. It has a
fairly wide range of statistical and graphical
capabilities and is relatively easy to learn to use.
PROGRAM DESCRIPTION
A data handling programming language which
S-PLUS™ is available on PCs and computers running the
Unix operating system. It is extremely flexible
(as one would expect of a programming
language) but is not as easy to use as some of the
other statistics packages.
A package for Generalized Linear Interactive
GLIM™ Modeling written by the Royal Statistical
Society. It is available on number of different
computer types and is very flexible. It is written
by statisticians for statisticians and requires a fair
degree of statistical understanding and
sophistication.
A simple statistics package written exclusively
STATVIEW™ for the Apple Macintosh. It has fairly limited
range of statistical facilities and is not very
flexible, but is adequate for simple hypothesis
testing and is easy to learn to use.
A statistics package available only for the
DATADESK™ Apple Macintosh which makes very effective use
of the Macintosh desktop interface. It has a
moderate range of statistical tools and good
dynamic graphics for data exploration. It is easy
to learn for anyone familiar with the Macintosh
interface.
A statistics software developed to
MICROSTAT II™ perform basic statistical tests for smaller
number of data. It is menu driven and
preferably a good software that one can
start with for basic statistical analysis.

Source: http://www.wsdinc.com/products

Descriptive statistics, frequency and probability

distributions, scatterplots, anova, crosstabs,
hypothesis testing (means, proportions, and
variance), power, sample size determination, non-
parametric tests, simple, multiple and stepwise
regression, correlation analysis, time series,
canonical correlation, principal components,
discriminant, cluster, orthogonal factor analysis
Provides statistical analysis software to the
NCSS 2001 occasional user of statistics. Comprehensive and
accurate. Includes over 200 procedures and
graphics.

Source: http:///www.ncss.com
PROGRAM DESCRIPTION
A statistics software developed by Statsoft,
STATISTICA™ Inc. and runs on Windows system.

Source: Statsoft, Inc. (1995), STATISTICA for

Windows

Frequency tables, crosstabulations,

correlation analysis, t tests, differences between
variances, chi-square tests, percentiles,
measures of central tendency, analysis of
variance (repeated measures, post-hoc tests test
of assumptions), canonical discriminant
analysis, multiple regression, non-linear
estimation, time-series forecasting, cluster
analysis, factor analysis, canonical correlation,
multidimensional scaling, process analysis,
experimental designs, loglinear analysis,
survival analysis.
A computer program that is useful
STATA for data analysis and manipulating data
sets. It is generally good with cross-
section data, but poor for time-series.

Source: qed.econ.queensu.ca/pub/computer
/sware/stat.,html
A statistics package available for
STATGRAPHICS™ PC only which combines statistics and
graphics for data analysis. It is fairly
easy to use and has reasonably wide
range of facilities.
A complete language for the
TSP™ estimation and simulation of
econometric models.

Source: http://www.tspintl.com

Regression analysis with first order serial

correlation, ordinary least squares, two-stage
least squares, hypothesis testing, time series
methods such as Box-Jenkins and Kalman filter
estimation, econometric problem solving
functions
Terms and Concepts for Review

Data analysis Questionnaire

Fixed-alternative Question Reliability
Non-probability Sampling Research Design
Observation Sampling
Open-ended Question Secondary Data Collection
Primary Data Collection Survey
Probability Sampling Validity

Assessment Tasks

1. Define research design in your own words.

2. Give an example for each type of research design.
3. Differentiate probability sampling from non-probability sampling. Give examples of
researches which utilized specific sampling techniques under each category.
4. Using Appendix A, Appendix B, and Appendix C, select the final samples through simple random
sampling using a sample size based on 90% reliability and a sampling error of 0.10.
5. Compare the advantages of using questionnaire rather than interview in a survey.
6. Identify the appropriate statistical tool that you will use for your study.