Key takeaways
– A simple random sample (SRS) is a subset of a population in which every member has an equal probability of selection.
– SRS minimizes selection bias and supports valid statistical inference, but it still has sampling error and can be impractical for large or hard‑to‑frame populations.
– Conducting an SRS requires: (1) a clearly defined population and sampling frame, (2) an appropriate sample size, (3) unbiased random selection (lottery, random numbers, or computer), and (4) careful handling of nonresponse and coverage issues.
– When an SRS is impractical or inefficient, alternatives include stratified, systematic, or cluster sampling.
Source: Investopedia — “Simple Random Sample” (Madelyn Goodnight).
What is a simple random sample?
A simple random sample (SRS) is a sample drawn from a population such that each possible member of the population has the same chance of being selected. When correctly implemented, an SRS is an unbiased basis for estimating population characteristics (means, proportions, totals) and for constructing confidence intervals and hypothesis tests.
Why use an SRS?
– Straightforward and easy to explain.
– Supports standard statistical formulas for variance, standard error, and confidence intervals.
– Appropriate when you have a complete sampling frame (list of population units) and want to avoid selection bias.
Room for error (sampling error and other considerations)
– Sampling error: Estimates from an SRS are probabilistic and have a margin of error (±). Larger samples reduce sampling error but cannot eliminate it.
– Non-sampling error: Coverage error (incomplete frame), nonresponse bias, measurement error, and data-entry mistakes can all bias results even if the sample was randomly selected.
– Practical constraints: Cost, time, and feasibility may limit sample size or the ability to obtain a true SRS.
How to conduct a simple random sample — six steps (with practical notes)
Step 1: Define the population
– Specify clearly who or what you want to study (e.g., “all full‑time employees of Company X as of Jan 1, 2025”).
– The population definition drives the sampling frame; be precise to avoid coverage errors.
Step 2: Choose the sample size
– Consider the desired margin of error (E), confidence level (Z), and expected variability (for proportions, use p; for means, use estimated standard deviation σ).
– For estimating a proportion, a common formula (large-sample approximation) is:
n0 = (Z^2 * p * (1 − p)) / E^2
• Z = 1.96 for 95% confidence, 2.576 for 99%.
• p is estimated proportion (use 0.5 if unknown to maximize required n).
– If population size N is not very large, apply the finite population correction:
n = n0 / (1 + (n0 − 1)/N)
– Example: Population N = 1,000, desired E = 3%, 95% confidence, p = 0.5:
n0 ≈ 1,067 → finite correction → n ≈ 517.
– For means, replace p(1−p) with σ^2 and use n0 = (Z^2 * σ^2) / E^2. Estimate σ from pilot data or literature.
Step 3: Determine population units (create the sampling frame)
– Build or obtain a complete list of all units (names, IDs, account numbers, inventory SKUs, etc.).
– A good sampling frame should match the defined population and avoid duplicates or omissions.
Step 4: Assign numerical values (or unique IDs) to units
– Give each unit a unique number (1..N). The numbering method doesn’t matter for randomness; consistency does.
– Record the mapping between numbers and units securely.
Step 5: Select random values (random selection)
Options:
• Lottery method: Put numbered slips in a hat — practical only for small N.
• Random-number tables: Traditional option using published random digits.
• Computer-generated random numbers: Use statistical software or spreadsheet functions (e.g., Excel’s RAND() with SAMPLE, R sample(), Python numpy.random.choice). Save the random seed for reproducibility.
• True-random services: random.org for entropy from atmospheric noise (rarely necessary).
Practical note: Avoid “human-chosen” numbers (birthdays, anniversaries) — these are not random.
Step 6: Identify the sample and collect data
– Map selected random numbers back to units and collect data from those units.
– Track response rates and reasons for nonresponse; follow up or replace units according to a predetermined protocol (avoid ad-hoc replacements that break randomness).
Random sampling techniques (brief)
– Simple random sampling (SRS): equal-probability selection of each unit.
– Variants: with replacement (each draw independent) or without replacement (common in practice). Most practical SRSs are without replacement.
Practical comparison: Simple random vs. other common methods
Simple Random vs. Stratified Random Sample
– Stratified sampling: population divided into subgroups (strata) and random samples taken within each stratum. Use when population is heterogeneous and you want precision within subgroups (e.g., by region, industry, age). Stratification improves precision if strata are internally homogeneous.
Simple Random vs. Systematic Sampling
– Systematic sampling: choose every kth unit after a random start. Easier to implement but can be biased if there is periodicity in the frame. If the list order is random (or unrelated to the variable of interest), systematic ≈ SRS in performance.
Simple Random vs. Cluster Sampling
– Cluster sampling: population divided into clusters (e.g., schools, branches); clusters are randomly selected and then all or a sample of units within selected clusters are measured. Useful for cost reduction and logistics, but intra‑cluster correlation reduces precision and requires cluster-adjusted variance estimation.
Advantages of a simple random sample
– Conceptually simple and unbiased if sampling frame is complete.
– Straightforward variance/margin-of-error formulas.
– No need for weighting if selection probabilities are equal and nonresponse is random.
Disadvantages of a simple random sample
– Requires a complete and accurate sampling frame — often unavailable.
– Inefficient if population is heterogeneous on key variables (stratification may be better).
– Can be costly or logistically difficult for dispersed populations (cluster sampling may be preferable).
– Nonresponse or coverage errors can invalidate the advantages of random selection.
Common pitfalls and how to avoid them
– Incomplete sampling frame: Invest in data cleaning or consider alternative designs.
– Nonresponse bias: Track response rates, perform follow-ups, and consider weighting or imputation.
– Ad-hoc replacements: Never replace nonrespondents with conveniently available units without adjusting analysis; replace according to a pre-specified random procedure.
– Failure to record the random seed or method: Document process for reproducibility and auditing.
Practical checklist for implementing an SRS
– Define population precisely.
– Build and validate the sampling frame.
– Decide sample size using desired precision and confidence (apply finite-population correction if needed).
– Assign unique IDs to every frame unit.
– Choose a reproducible random selection method and save the seed or algorithm.
– Collect data and monitor response rates, documenting any deviations.
– Compute estimates with margins of error and report design and nonresponse considerations.
Examples
– Simple lottery: Choose 25 employee names out of a hat from a company with 250 employees; each employee has a 1/10 chance.
– S&P 500 example: If you want a random subset of 20 companies from the S&P 500, assign numbers 1–500 and use a computer random-number generator to pick 20 unique numbers. Map those back to companies.
– Survey example: High school (N = 1,000), sample n = 100, observed 8 left-handed (p̂ = 0.08). The standard error = sqrt(p̂(1−p̂)/n) ≈ 0.027, so a 95% margin ≈ 1.96*0.027 ≈ 0.053 (±5.3%). So estimate = 8% ± 5.3%.
Tools and software (practical)
– Spreadsheets: Excel — use RAND() to assign random numbers and sort, or use the Data Analysis add-ins / RANDBETWEEN / new SAMPLE function in newer Excel.
– R: sample(x, size, replace = FALSE) — set.seed() for reproducibility.
– Python: numpy.random.choice(array, size, replace=False) — set numpy.random.seed().
– Random.org: True random numbers for special cases.
When to choose alternatives to SRS
– If you lack a complete list of the population → consider cluster sampling.
– If key subgroups are small but important → use stratified sampling.
– If units are naturally ordered and you want simplicity and reduced cost → systematic sampling (check for periodicity).
How random samples are used in finance and research
– Market research surveys, opinion polling, randomized controlled trials, audit sampling, portfolio performance studies (randomly selecting securities for deeper analysis), regulatory compliance testing. Random samples allow inference from a manageable number of observations to a larger population.
The bottom line
A simple random sample is an unbiased, easy‑to‑explain method for selecting units from a population when a complete sampling frame is available. It supports standard inferential statistics, but researchers must plan sample size, guard against nonresponse and coverage problems, and consider alternatives (stratified, systematic, cluster) when SRS is impractical or inefficient.
Further reading / source
– Investopedia, “Simple Random Sample” (Madelyn Goodnight).
Editor’s note: The following topics are reserved for upcoming updates and will be expanded with detailed examples and datasets.