Lecture Notes
Coursework in Business Statistics
Data Collection Methods and Sampling
R. Sivasamy
April 26, 2026
Contents
1 Methods of Data Collection and Sampling Methods 2
1.1 Methods of Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Direct Observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Personal Interview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Experimentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.4 Documentary Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Questionnaire Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Population, Parameters, Census and Sampling 3
2.1 Advantages and Disadvantages of Sampling . . . . . . . . . . . . . . . . . . . 4
2.2 When and Why We Conduct a Census . . . . . . . . . . . . . . . . . . . . . 4
2.3 Sampling Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3.1 Probability Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3.2 Non-Probability Sampling . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Marketing Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.5 Case Study: Marketing Research Survey on Consumer Behaviour . . . . . . 5
2.6 Sample Size Determination for Proportions . . . . . . . . . . . . . . . . . . . 7
2.7 Proportional Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1
1 Methods of Data Collection and Sampling Methods
1.1 Methods of Data Collection
In Business Statistics, the quality of statistical conclusions depends critically on the method
used to collect data. The following are the major methods of primary and secondary data
collection.
1.1.1 Direct Observation
Direct observation involves collecting data by watching events, behaviours, or conditions as
they occur. Advantages: Simple, inexpensive, and free from respondent bias. Limita-
tions: Only suitable for observable phenomena; may be time-consuming.
1.1.2 Personal Interview
In this method, the investigator interacts directly with respondents to obtain informa-
tion. Advantages: High response rate, clarification possible. Limitations: Costly, time-
intensive, and subject to interviewer bias.
1.1.3 Experimentation
Experimentation involves manipulating one or more variables under controlled conditions
to study cause–effect relationships. Applications: Product testing, pricing experiments,
advertising effectiveness. Limitations: Expensive and sometimes unrealistic outside labo-
ratory settings.
1.1.4 Documentary Evidence
Documentary evidence refers to the use of existing records such as company reports, govern-
ment publications, financial statements, and archival data. Advantages: Low cost, readily
available. Limitations: Data may be outdated or collected for a different purpose.
1.2 Questionnaire Design
A questionnaire is a structured set of questions used to collect information from respondents.
Effective questionnaire design includes:
• Clear objectives aligned with the research problem.
2
• Simple, unambiguous, and unbiased questions.
• Logical sequencing: demographic questions last.
• Appropriate question types: open-ended, closed-ended, Likert scales.
• Pilot testing to identify errors and improve clarity.
Example of a Questionnaire The following is an example of a short questionnaire de-
signed to study consumer behaviour toward a new beverage product. The questionnaire
begins with a brief introduction stating the purpose of the survey and assuring respondents
of confidentiality. It then includes basic demographic items such as age group and gender,
followed by product-related questions.
Respondents are asked to indicate the frequency of purchasing soft drinks (e.g., “Daily”,
“Weekly”, “Occasionally”), their preferred type of beverage, and their level of agreement
with statements such as “The price of the product influences my purchase decision” using
a five-point Likert scale ranging from “Strongly Disagree” to “Strongly Agree”. Additional
questions assess brand awareness (e.g., “Have you heard of the new product?” with “Yes/No”
options) and purchase intention (e.g., “How likely are you to buy this product?” with a scale
from 1 to 5). The questionnaire concludes with an open-ended question inviting respondents
to provide suggestions for improving the product. This structure ensures clarity, relevance,
and ease of analysis for introductory marketing research applications.
2 Population, Parameters, Census and Sampling
In Business Statistics, a population refers to the complete set of individuals, items, or ob-
servations relevant to a study, while a parameter is a numerical measure that describes a
characteristic of the entire population, such as the population mean or proportion.
A census involves collecting data from every unit in the population, whereas sampling in-
volves selecting only a subset of units for study. A statistic is a numerical measure computed
from sample data, such as the sample mean or sample proportion, and is used to estimate
the corresponding population parameter.
The key difference between a parameter and a statistic is that a parameter is fixed but
usually unknown, while a statistic varies from sample to sample and is observable. Similarly,
a census provides complete information but is often costly and time-consuming, whereas a
sample survey is quicker, less expensive, and more practical for large populations.
3
2.1 Advantages and Disadvantages of Sampling
Sampling offers several advantages in statistical investigations. It is generally less costly and
less time-consuming than conducting a full census, making it practical for large populations.
Sampling also allows for quicker data collection, easier supervision of fieldwork, and often
yields more accurate results because trained investigators can focus on a smaller number of
units.
However, sampling has limitations: results are subject to sampling error, and biased
conclusions may arise if the sample is not representative of the population. In addition,
sampling may be inappropriate when the population is very small or when extremely high
accuracy is required, in which case a census may be more suitable.
2.2 When and Why We Conduct a Census
A census is conducted when information is required from every unit in the population,
typically in situations where the population is small, highly important, or where extremely
high accuracy is essential.
Governments often conduct national censuses to obtain complete demographic and socio-
economic data for planning and policy formulation. Organisations may also conduct a census
when the cost and time involved are manageable or when even small errors could lead to
incorrect decisions.
The main reason for conducting a census is that it provides comprehensive, exact infor-
mation without sampling error, ensuring full coverage of the population.
2.3 Sampling Methods
Sampling is the process of selecting a subset of units from a population to draw valid con-
clusions.
2.3.1 Probability Sampling
Every unit in the population has a known, non-zero probability of selection. Common types
include:
• Simple Random Sampling (SRS): Each unit has an equal chance of selection.
• Systematic Sampling: Selecting every k-th unit after a random start.
• Stratified Sampling: Dividing the population into homogeneous strata and sampling
from each.
4
• Cluster Sampling: Selecting entire groups (clusters) randomly.
2.3.2 Non-Probability Sampling
Selection is based on judgement, convenience, or other non-random criteria. Types include:
• Convenience Sampling: Selecting units that are easily accessible.
• Judgement (Purposive) Sampling: Researcher selects units based on expertise.
• Quota Sampling: Ensuring representation of key groups without random selection.
• Snowball Sampling: Existing respondents recruit future respondents.
2.4 Marketing Research
Marketing research is the systematic process of collecting, analysing, and interpreting data
to support marketing decisions. Key areas: consumer behaviour, product development,
pricing, promotion, and distribution strategies.
2.5 Case Study: Marketing Research Survey on Consumer Be-
haviour
A retail company wishes to understand consumer preferences for a new product line. The
steps include:
1. Problem Definition: Identify factors influencing consumer purchase decisions.
2. Research Design: Use a structured questionnaire with Likert-scale items.
3. Sampling Plan: Apply stratified sampling based on age groups.
4. Data Collection: Administer online surveys and in-store interviews.
5. Analysis: Summarise preferences, compute proportions, and test hypotheses.
6. Reporting: Provide recommendations for product positioning and marketing strategy.
5
Example of Research Design The study adopts a descriptive research design using a
structured questionnaire consisting of both closed-ended and Likert-scale items. The ques-
tionnaire is organised into sections covering demographic characteristics, purchasing be-
haviour, and attitudinal measures related to the product. Respondents indicate their level
of agreement with statements on a five-point Likert scale ranging from “Strongly Disagree”
(1) to “Strongly Agree” (5). This design ensures standardised data collection, facilitates
quantitative analysis, and allows the researcher to examine patterns in consumer behaviour
with respect to price sensitivity, taste preference, and purchase intention.
Case-Study Dataset for the Consumer Behaviour Questionnaire A sample dataset
was constructed based on responses collected from 420 consumers who completed the ques-
tionnaire on purchasing behaviour for a new beverage product. The dataset contains both
demographic and behavioural variables aligned with the questionnaire items. Demographic
variables include AgeGroup (18–25, 26–35, 36–50, 51+), Gender (Male, Female, Other), and
IncomeLevel (Low, Middle, High). Behavioural variables capture purchasing patterns such
as PurchaseFrequency (Daily, Weekly, Occasionally), PreferredBeverageType (Soft drinks,
Juices, Energy drinks, Water), and BrandAwareness (Yes/No). Attitudinal variables are
recorded using five-point Likert scales, including PriceInfluence, TastePreference, and Health-
Concern, where higher scores indicate stronger agreement.
The dataset also includes PurchaseIntention, measured on a 1–5 scale, and an open-ended
text field summarised as ConsumerSuggestions. This case-study dataset provides a realistic
foundation for teaching data summarisation, graphical analysis, estimation of proportions,
and introductory hypothesis testing in Business Statistics.
Measurement on a 1–5 Scale Variables such as PriceInfluence, TastePreference, and
HealthConcern are measured on a 1–5 scale, where respondents indicate their level of agree-
ment or intensity of preference. A value of 1 typically represents the lowest level (e.g.,
“Strongly Disagree” or “Very Low”), while a value of 5 represents the highest level (e.g.,
“Strongly Agree” or “Very High”). Intermediate values (2, 3, and 4) capture increasing
degrees of agreement or preference, allowing the researcher to quantify attitudes and analyse
patterns in consumer behaviour.
Reporting The final stage of the study involves preparing a concise report that sum-
marises the key findings and translates them into actionable recommendations for prod-
uct positioning and marketing strategy. Based on the analysed data, the report highlights
consumer preferences, price sensitivity, and purchase intention, and uses these insights to
6
propose strategies such as emphasising taste attributes, adjusting price points, or targeting
specific demographic segments. The recommendations guide decision-makers in positioning
the product effectively in the market and designing promotional activities that align with
consumer behaviour patterns.
2.6 Sample Size Determination for Proportions
When estimating a population proportion p with a specified margin of error E and confidence
level (1 − α), the required sample size is:
2
Zα/2 p(1 − p)
n= .
E2
If no prior estimate of p is available, use p = 0.5 to obtain the maximum required sample
size:
2
Zα/2 × 0.25
n= .
E2
Example: To estimate a consumer preference proportion with 95% confidence (Z0.025 =
1.96) and margin of error E = 0.05:
1.962 × 0.25
n= = 384.16 ≈ 385.
0.052
Thus, a sample of at least 385 respondents is required.
2.7 Proportional Allocation
Proportional allocation is a sampling technique used in stratified sampling where the sample
size drawn from each stratum is proportional to the size of that stratum in the population.
If the population consists of strata with sizes N1 , N2 , . . . , Nh and the total sample size is n,
then the sample from stratum i is given by ni = n NNi , where N is the total population
size.
For example, suppose a population of N = 1,000 consumers is divided into three age
groups: 18–25 (N1 = 300), 26–40 (N2 = 500), and 41+ (N3 = 200). If a sample of n = 100
respondents is required, proportional allocation yields n1 = 30, n2 = 50, and n3 = 20. This
ensures that the sample reflects the population structure accurately.