SAMPLE & SAMPLING

Detailed Study Notes for Examinations

Research Methodology | Statistics | Public Health

Advantages &
Definition of Sample Types of Sample Sampling Techniques
Disadvantages

Prepared for Academic / Examination Use

Page 1
 SECTION 1 — DEFINITION OF SAMPLE

1.1 What is a Sample?

Definition: A sample is a subset or representative portion selected from a larger group called the
population (or universe). It is chosen in such a way that it faithfully reflects the characteristics of the
whole population, so that conclusions drawn from the sample can be validly generalised to the
population from which it was drawn.

The population refers to the entire collection of individuals, objects, events, or measurements that share at
least one common characteristic and are of interest to the researcher. Because studying an entire
population is often impractical, costly, or impossible, researchers select a sample and use statistical
inference to make conclusions about the population.

Key Terminologies:
Term Meaning

Population (Universe) The complete set of all elements of interest in a study.

Sample A subset of the population selected for actual study.

Sampling Unit The individual element or group of elements that constitutes a single unit for
sampling.

Sampling Frame A complete list or map of all units in the population from which a sample is
drawn.

Parameter A numerical characteristic of a population (e.g., population mean µ).

Statistic A numerical characteristic of a sample (e.g., sample mean x■).

Census Study of every member of the population — the opposite of sampling.

SECTION 2 — TYPES OF SAMPLE

Samples can be broadly categorised based on the method of selection and the nature of the study. The
two primary classifications are Probability Samples and Non-Probability Samples.

2.1 Probability Samples (Random Samples)

In probability sampling, every element of the population has a known and non-zero probability of
being selected. This allows for statistical inference and generalisation.

Type Description When to Use

Simple Random Every member has an equal chance of selection. Homogeneous population;
Sample (SRS) Done via lottery or random number tables. complete sampling frame
available.

Page 2
 Systematic Random Select every k-th element from a list after a Ordered lists, production
Sample random start. k = N/n (sampling interval). lines, patient registers.

Stratified Random Population divided into homogeneous subgroups Heterogeneous population

Sample (strata); random samples drawn from each with clear subgroups (age,
stratum. gender, region).

Cluster (Area) Population divided into clusters (e.g., villages); Large geographical areas; no
Sample entire clusters are randomly selected. complete sampling frame
available.

Multi-stage Sample Sampling done in multiple stages — clusters National surveys,

selected, then sub-samples within clusters. epidemiological studies over
wide areas.

Multi-phase Sample Information collected in phases; sub-sample from Health surveys requiring
first phase is studied in greater depth. detailed follow-up on a
subset.

2.2 Non-Probability Samples (Non-Random Samples)

In non-probability sampling, the probability of each element being selected is unknown. Results
cannot be statistically generalised to the whole population.

Type Description Limitation

Convenience Units selected on the basis of easy availability or Highly prone to bias; not
(Accidental) Sample accessibility. representative.

Purposive Researcher deliberately selects units based on Subjective; depends entirely

(Judgmental) knowledge and judgment. on researcher expertise.
Sample

Quota Sample Population divided into groups; researcher fills a Selection within quota is
fixed quota from each group — but non-randomly. subjective; may not be
representative.

Snowball Sample Initial subjects recruit further subjects from their Sampling bias;
networks. Used for hidden or hard-to-reach network-dependent; not
populations. random.

Volunteer Individuals volunteer to participate. Common in Strong response bias;

(Self-Selection) online surveys. over-representation of
Sample motivated individuals.

Page 3
 SECTION 3 — DETAILED NOTES ON SAMPLING

3.1 Definition of Sampling

Sampling is defined as the process or technique of selecting a representative subset (sample)
from a larger population with the objective of studying characteristics of the population, drawing
valid inferences, and making decisions — without examining every member of the population. It is a
fundamental tool in research, epidemiology, public health, statistics, and social sciences.

According to Cochran (1977): 'Sampling is the process of selecting a part of the population for the
purpose of studying it and drawing conclusions about the whole.'

According to Goode & Hatt: 'A sample is a smaller representation of a larger whole.'

3.2 Essence / Need / Rationale for Sampling

The need for sampling arises because studying every element of a large population is often impractical.
The essence of sampling lies in the following reasons:

• Economy of Resources: It is far less expensive to study a sample than the entire population. Data
collection costs in time, money, and personnel are drastically reduced.
• Time Efficiency: Sampling yields results much faster than a census, which is crucial in emergency
public health situations such as disease outbreaks.
• Practicability: For infinite, very large, or geographically dispersed populations, a census is physically
impossible. Sampling provides a workable alternative.
• Greater Accuracy: Paradoxically, a well-designed sample may yield MORE accurate data than a
census because fewer personnel are needed, reducing non-sampling errors (e.g., data entry mistakes,
interviewer fatigue).
• Destructive Testing: When testing destroys the unit under study (e.g., drug potency testing, blood
testing), only sampling is feasible.
• Inaccessible Populations: Some populations (e.g., endangered species, hard-to-reach communities,
rare diseases) are too small or scattered for a full census.
• Scientific Inference: Probability sampling allows the use of statistical theory to estimate population
parameters with measurable degrees of precision and confidence.

3.3 Basic Principles / Characteristics of Good Sampling

• Representativeness – The sample must accurately reflect the composition of the population.
• Adequacy – The sample size must be large enough to ensure reliable conclusions.
• Randomness – Each element of the population should have a known probability of selection (for
probability sampling).
• Homogeneity – Units within each stratum (if stratified) should be as similar as possible.
• Feasibility – The sampling procedure must be practical, affordable, and executable.
• Precision – The sample should produce estimates with a small margin of error.

SECTION 4 — SAMPLING TECHNIQUES (METHODS)

Sampling techniques are broadly classified into two major categories:

Page 4
 Category Key Feature Examples

A. Probability Sampling Every unit has a known, non-zero SRS, Systematic,

(Random Sampling) probability of selection. Statistical inference Stratified, Cluster,
is valid. Multi-stage

B. Non-Probability Sampling Selection is based on judgment, Convenience, Purposive,

(Non-Random Sampling) convenience, or availability. Cannot Quota, Snowball
generalise statistically.

A. PROBABILITY SAMPLING TECHNIQUES

1. Simple Random Sampling (SRS)

Definition: Every member of the population has an equal and independent chance of being
selected. Selection is done using a lottery method or a random number table.

Procedure:
• Number all elements from 1 to N.
• Choose n numbers randomly (via lottery or random number generator).
• The elements corresponding to chosen numbers form the sample.
Types: (i) With replacement (SRSWR) — selected unit returned before next draw. (ii) Without replacement
(SRSWOR) — selected unit not returned.
Formula for Sample Size: n = Z2 × p × q / e2 (where Z = Z-value at desired confidence, p = expected
proportion, q = 1-p, e = acceptable error)

2. Systematic Random Sampling

Definition: Select every k-th element from a list after choosing a random starting point between 1
and k. The sampling interval k = N/n.

Example: Population N = 1000, desired sample n = 100. k = 1000/100 = 10. Choose random start (e.g.,
7). Select elements 7, 17, 27, 37 … 997.
Precaution: Avoid systematic bias (periodicity) — if the list has a cyclical pattern matching k, the sample
will be biased.

3. Stratified Random Sampling

Definition: The population is divided into non-overlapping, homogeneous subgroups called

strata. An independent random sample is then drawn from each stratum.

Types of Allocation:

Allocation Type Formula / Rule

Proportional Allocation n_h = n × (N_h / N) — strata sampled in proportion to their size.

Optimum (Neyman) Allocation n_h ∝ N_h × S_h — larger or more variable strata get bigger samples.

Equal Allocation Same number drawn from each stratum regardless of size.

Page 5
Best for: Studies where the population has distinct sub-groups relevant to the research variable (e.g.,
age-stratified health surveys).

4. Cluster (Area) Sampling

Definition: The population is divided into clusters (naturally occurring groups, e.g., villages, wards,
households). A random sample of clusters is selected, and all members of the chosen clusters are
studied.

Difference from Stratified: In stratified sampling, we sample FROM each stratum (within-group). In
cluster sampling, we sample ENTIRE clusters (between-group). Clusters should be as heterogeneous
internally (diverse within) as possible.
Use: National immunisation surveys (WHO EPI cluster sampling — 30×7 method), household surveys in
rural areas.

5. Multi-stage Sampling

Definition: Sampling is carried out in two or more successive stages. At each stage, a random
sample is drawn from the units selected in the previous stage.

Example (3-stage): Stage 1 — randomly select districts from a country; Stage 2 — randomly select
villages from chosen districts; Stage 3 — randomly select households from chosen villages.

B. NON-PROBABILITY SAMPLING TECHNIQUES

1. Convenience (Accidental) Sampling
Subjects are selected because they are easily accessible to the researcher (e.g., patients in a waiting
room, students in a class). Quick and cheap but highly susceptible to selection bias.

2. Purposive (Judgmental) Sampling

The researcher deliberately selects specific units based on his/her expert judgment that they are
representative or best serve the research purpose. Common in qualitative research and case studies.

3. Quota Sampling
The population is divided into subgroups, and the researcher fills a pre-determined quota from each
subgroup — but selection within the quota is NOT random. Similar in structure to stratified sampling but
without randomisation.

4. Snowball Sampling
Initial subjects are identified and then asked to recruit additional subjects from their social networks.
Particularly useful for studying hidden or stigmatised populations (e.g., drug users, sex workers, rare
disease patients).

5. Volunteer / Self-selection Sampling

Participants volunteer themselves for the study (e.g., online surveys). Carries a high risk of response bias
since volunteers may differ systematically from non-volunteers.

Page 6
 SECTION 5 — ADVANTAGES AND DISADVANTAGES OF SAMPLING

5.1 Advantages of Sampling

Advantage Explanation

✓ Economy Collecting data from a sample costs a fraction of what a census would require.
Fewer resources — money, personnel, and equipment — are needed.

✓ Speed / Sampling allows data collection, analysis, and reporting to be completed much faster
Time-saving than a census — vital in emergency public health responses.

✓ Greater Accuracy A smaller study means fewer enumerators, less supervision difficulty, and fewer
non-sampling errors. A well-conducted sample survey can be more accurate than a
poorly conducted census.

✓ Feasibility For very large, infinite, or geographically dispersed populations, sampling is the only
practical approach.

✓ Destructive When the act of measurement destroys the unit (blood tests, drug assays, product
Testing quality control), sampling is the only option.

✓ Access to Detailed Because fewer subjects are studied, researchers can collect more detailed and
Information in-depth information from each sampled unit.

✓ Statistical Probability sampling allows researchers to estimate population parameters with

Precision confidence intervals, quantifying the degree of certainty.

✓ Administrative Sampling frames, logistics, data management, and quality control are all easier to
Convenience manage with a sample than with an entire population.

5.2 Disadvantages / Limitations of Sampling

Disadvantage Explanation

✗ Sampling Error Since only a subset is studied, the sample statistic may differ from the true
population parameter. This difference is called sampling error. Larger samples
reduce this error.

✗ Bias Risk If the sample is not properly selected (non-representative or skewed selection),
results will be biased and misleading. Non-probability samples are especially
vulnerable.

✗ Requirement of Designing a valid sampling plan demands statistical knowledge. Incorrect choice
Expertise of sample size or method can invalidate results entirely.

✗ Non-applicability in When the population itself is very small, a census is more appropriate. Sampling
Small Populations becomes impractical when N is too small to allow meaningful subset selection.

✗ Incomplete Sampling A biased or outdated sampling frame (list of population units) can systematically
Frame exclude segments of the population, introducing frame error.

Page 7
✗ Non-sampling Errors Errors from respondent non-response, measurement inaccuracies, interviewer
bias, or data entry mistakes can occur and may be harder to detect in a sample
study.

✗ Legal / Administrative Some contexts legally require complete enumeration (e.g., national census for
Requirements electoral delimitation, certain clinical trials), making sampling insufficient.

✗ Heterogeneous When a population is extremely diverse, selecting a truly representative sample

Populations is complex and may require sophisticated stratified or cluster designs.

SECTION 6 — SAMPLING ERRORS vs. NON-SAMPLING ERRORS

Aspect Sampling Error Non-Sampling Error

Definition Difference between sample statistic Errors arising from sources other
and true population parameter. than the sampling process itself.

Cause Due to chance variation in selecting Due to measurement, recording,

a sample. interviewer bias, non-response.

Occurrence Only in sample surveys. Occurs in both census and sample

surveys.

Control Reduced by increasing sample size. Reduced by training,

standardisation, careful design.

Measurability Can be statistically estimated Difficult to measure or quantify.

(standard error).

Examples Random fluctuation in mean blood Recording error, misunderstanding

pressure estimate. of questions, data entry mistakes.

SECTION 7 — QUICK REVISION / EXAM TIPS

• Tip 1: Sample = subset; Population = whole. Statistic describes the sample; Parameter describes the
population.
• Tip 2: Probability sampling → random → generalisable. Non-probability sampling → non-random → not
generalisable.
• Tip 3: Stratified: divide into strata, sample WITHIN each stratum. Cluster: divide into clusters, sample
ENTIRE clusters.
• Tip 4: Systematic sampling interval k = N/n. Always choose random start point first.
• Tip 5: Snowball sampling is best for hidden/hard-to-reach populations.
• Tip 6: Sampling error ↓ as sample size ↑. Non-sampling errors can occur even in a census.
• Tip 7: 30×7 cluster method (WHO/EPI): 30 clusters × 7 children = 210 subjects — used for vaccination
coverage surveys.
• Tip 8: Optimum (Neyman) allocation: allocate more to strata with greater variability (higher SD) or larger
size.
• Tip 9: Remember: A small well-designed sample > a large poorly-designed census in terms of
accuracy.

Page 8
• Tip 10: Always state both advantages AND disadvantages of sampling in exam — examiners expect
balanced answers.

End of Notes — Sample & Sampling | Prepared for Examination Use | All rights reserved

Page 9