BASIC DEFINITIONS ON SAMPLING:

1. Determining the Universe

The universe (also called the population) in research refers to the entire group of individuals, objects,
or events about which the researcher intends to draw conclusions. It represents all possible elements
that fit the scope of the study. Determining the universe is the first and most critical step in the
sampling process because an unclear definition can lead to inaccurate or misleading results. For
example, if a company wants to study customer satisfaction with its e-commerce platform, the
universe could be defined as all registered customers who made at least one purchase in the last six
months. Clearly defining the universe helps in setting boundaries for data collection, ensuring that
the study addresses the right target group.

2. Sampling Frame

A sampling frame is the actual list or database of elements from which the sample will be drawn. It
acts as a bridge between the defined universe and the selected sample. A good sampling frame
should be complete, accurate, and up-to-date. For instance, if a retail chain wants to conduct
research on customer buying behavior, its sampling frame could be the loyalty program database,
which contains names, contact details, and purchase history of customers. However, if this list misses
customers who are not loyalty members, the study may suffer from frame error, as some elements of
the universe are excluded. Thus, choosing a reliable and representative sampling frame is vital for
minimizing bias in research outcomes.

3. Sampling Unit

The sampling unit is the smallest and most basic element from which information is collected during
sampling. It represents the specific entity chosen for study within the sampling frame. Sampling units
can vary depending on the research problem—they may be individuals, households, organizations, or
even transactions. For example, if a business researcher is studying the consumption patterns of
packaged foods, the sampling unit could be the household. In another case, if a bank wants to study
the efficiency of loan processing, the sampling unit might be an individual loan application. The
correct identification of sampling units ensures that the data collected directly relates to the research
objectives and avoids inconsistencies in analysis.

PROBABILITY & NON-PROBABILTY SAMPLING METHODS

1. Probability and Non-Probability Sampling Methods

Sampling is a vital step in business research as it helps in selecting a representative portion of the
population for study. Broadly, sampling methods are classified into probability and non-probability
approaches. In probability sampling, every element of the population has a known and non-zero
chance of being selected. This ensures representativeness and allows statistical generalization. For
example, a company wanting to study employee satisfaction in all its branches may use probability
sampling to ensure unbiased selection. On the other hand, non-probability sampling does not
provide equal chances of selection, often relying on the researcher’s judgment or accessibility. While
it is less costly and faster, it carries a higher risk of bias and limited generalizability. For example, a
retail store conducting a quick survey among customers visiting on a particular day is using non-
probability sampling.

2. Probability Method – Simple Random Sampling

Simple random sampling is the most basic form of probability sampling, where every individual in the
population has an equal chance of selection. This can be done using lottery methods, random
number tables, or computer-generated randomization. For instance, if a bank wants to study
customer satisfaction across all account holders, it may randomly select 500 customers from its
database. The advantage of this method is its fairness and lack of bias. However, it may not be
practical for very large or geographically dispersed populations.

3. Probability Method – Systematic Sampling

In systematic sampling, researchers select every kth element from a list after choosing a random
starting point. For example, if a company wants to survey 1,000 customers from a database of
10,000, it might select every 10th customer after a random start. This method is simpler to
administer than random sampling and ensures a spread-out sample. However, if the population list
has a hidden pattern (such as customers listed by purchase frequency), it may introduce bias.

4. Probability Method – Stratified Random Sampling

Stratified random sampling divides the population into homogeneous subgroups (strata) such as age,
gender, income level, or region, and then random samples are drawn from each stratum. This
ensures representation of all critical segments. For example, if a multinational company is
researching employee engagement, it may stratify employees by department (finance, HR,
marketing, production) and then sample proportionally from each. This method increases precision
and ensures that minority groups are not overlooked.

5. Probability Method – Cluster Sampling

In cluster sampling, the population is divided into clusters (often based on geography or natural
groupings), and then entire clusters are randomly selected for study. For example, a fast-food chain
studying customer preferences across India might randomly select 10 cities (clusters) and survey all
or some customers within those cities. Cluster sampling is cost-effective and convenient, especially
for large populations spread across wide areas. However, since only a few clusters are studied, the
sample may be less representative compared to stratified sampling.

6. Non-Probability Sampling – Convenience Sampling

Convenience sampling relies on selecting respondents who are easily accessible to the researcher.
For example, a store manager collecting feedback from customers present in the outlet at the time of
visit is using convenience sampling. While it is inexpensive and quick, this method carries a high risk
of bias, as it may not represent the entire population. It is often used in pilot studies or exploratory
research where speed is more important than accuracy.

7. Non-Probability Sampling – Judgmental (Purposive) Sampling

In judgmental or purposive sampling, the researcher uses their expertise to select respondents who
are most likely to provide relevant information. For example, if a firm wants to understand the
challenges of international business expansion, it may purposively interview only senior managers
with overseas experience. This method is useful when specialized knowledge is required, but it is
subjective and lacks the objectivity of probability-based methods.

8. Non-Probability Sampling – Snowball Sampling

Snowball sampling is commonly used in cases where the population is difficult to identify or reach.
The process begins with a few known respondents, who then refer the researcher to additional
participants, and the sample grows like a snowball. For example, in business research on informal
entrepreneurs or high-net-worth investors, initial contacts may introduce the researcher to others
within their networks. While effective for hidden populations, snowball sampling often results in
biased samples, as participants are drawn from interconnected social circles.

9. Non-Probability Sampling – Quota Sampling

Quota sampling involves dividing the population into subgroups (like stratified sampling) but without
random selection. Instead, researchers select respondents until a quota for each subgroup is filled.
For instance, a consumer goods company studying brand preferences may set quotas for 40% male
and 60% female respondents, ensuring gender representation, but selection within each group may
depend on convenience. While it ensures subgroup representation, it is still prone to researcher bias
in respondent selection.
 Sampling and Non-Sampling Errors in Business Research

1. Sampling Errors
Definition:
Sampling error is the difference between a sample statistic and the actual population parameter,
arising due to using only a part (sample) instead of the whole (population).

Formula: Sampling Error = Sample Statistic − Population Parameter

Types of Sampling Errors

a. Random Sampling Error

Occurs due to natural chance or variability. Even well-designed samples can differ from the
population.

Example (Business): A bank surveys 50 loan customers and finds 80% satisfaction. However, the
actual population satisfaction might be 70%, showing a random sampling error.

b. Systematic Sampling Error

Caused by flaws in the sampling method. Certain groups are systematically excluded.

Example (Business): A food delivery company collects feedback only from mobile app users,
excluding website users. This skews satisfaction results.

How to Reduce Sampling Errors

• Increase sample size
• Use probability-based sampling (simple random, stratified, cluster)
• Ensure sample represents all segments

2. Non-Sampling Errors
Definition:
Non-sampling errors are errors not related to the act of sampling. They can occur during data
collection, processing, analysis, or interpretation—even in a census.

Types of Non-Sampling Errors

a. Measurement Error
Incorrect responses due to unclear questions or instruments.

Example: A survey asks, "Do you regularly use digital wallets?" but doesn't define "regularly."
Different users interpret it differently, leading to misleading data.

b. Non-Response Error
Occurs when some respondents fail to participate or skip questions.

Example: A retail brand emails a survey, but only highly satisfied customers respond. The results
appear overly positive.
c. Processing Error
Mistakes during data entry, coding, or analysis.

Example: A data analyst mistakenly enters ₹10000 as ₹1000, distorting income statistics in a
consumer study.

d. Interviewer Bias
When the interviewer’s tone, wording, or behavior influences the response.

Example: A sales executive conducting a product feedback survey praises the product before asking
questions, causing biased responses.

How to Reduce Non-Sampling Errors

• Pretest questionnaires
• Train data collectors
• Automate and validate data entry
• Use multiple channels to reach non-respondents
• Ensure clarity and neutrality in questions

Comparison Table
Aspect Sampling Error Non-Sampling Error Business Example
Occurs When Only a part of During any stage of Survey only urban
population is the research process customers vs.
surveyed skipping income
question
Cause Sample not Human error, biases, Sampling regionally
representative instrument error, etc. vs. unclear question
design
Reducible by Better sampling Better processes, Use stratified
methods training, validation sampling vs. pilot
survey
 Type I and Type II Errors in Hypothesis Testing

Introduction
Hypothesis testing is a fundamental part of statistical inference used in business decision-making. It
involves testing an assumption (the null hypothesis, H₀) against an alternative hypothesis (H₁). Two
types of errors can occur in this process: Type I and Type II errors.

Type I Error (False Positive)

A Type I Error occurs when the null hypothesis (H₀) is true, but we incorrectly reject it. This means we
detect an effect or difference when there is none.

• Also known as a 'False Positive'.

Business Example:
A bank introduces a new credit scoring algorithm. The null hypothesis (H₀) is that the new model
does not perform better than the existing model.
If the bank rejects H₀ and adopts the new model, believing it's better when in reality it is not, they
have made a Type I Error. This can lead to poor credit decisions, increasing risk and losses.

Type II Error (False Negative)

A Type II Error occurs when the null hypothesis (H₀) is false, but we fail to reject it. This means we
miss an actual effect or difference.

• Also known as a 'False Negative'.

Business Example:
A company tests a new marketing campaign. The null hypothesis (H₀) is that the new campaign is not
more effective than the current one.
If the company accepts H₀ and continues with the old campaign despite
the new one being better, they have made a Type II Error. This results in missed sales opportunities
and lower market share.

Comparison of Type I and Type II Errors

Aspect Type I Error Type II Error
Definition Rejecting H₀ when it is true Failing to reject H₀ when it is
false
Also Called False Positive False Negative
Business Impact Taking unnecessary action Missing a valuable
based on false detection opportunity
Conclusion
Understanding Type I and Type II errors helps businesses make more informed decisions. Reducing
these errors requires careful experimental design, appropriate sample sizes, and setting the right
level of significance.