Random Sampling Methods

When researchers want to learn something about a large group — called a population — it is usually impossible or impractical to collect data from every single member. Instead, they study a smaller subset called a sample. The critical challenge is making sure that sample accurately reflects the whole population. Random sampling methods solve this problem by using chance to select individuals, removing human bias from the process and giving every member of the population a known, non-zero probability of being chosen. Random sampling is the gold standard in statistics because it allows us to use probability theory to quantify how confident we are in our conclusions. When a sample is chosen randomly, the laws of probability guarantee that, on average, the sample will resemble the population — and crucially, we can calculate exactly how much error to expect. This is what makes it possible to poll 1,000 voters and reliably predict the preferences of millions, or to test 200 light bulbs and draw conclusions about an entire factory's output. There are several distinct random sampling techniques, each suited to different situations: simple random sampling, systematic sampling, stratified sampling, and cluster sampling. Choosing the right method depends on the structure of the population, the resources available, and the precision required. Each method guarantees randomness but achieves it differently, with different trade-offs in cost, complexity, and statistical efficiency.

Key takeaways
  • Simple Random Sampling (SRS)
  • Systematic Sampling
  • Stratified Sampling
  • Cluster Sampling

Simple Random Sampling (SRS)

In simple random sampling, every possible sample of size n drawn from a population of size N has an equal probability of being selected. In practice, each individual is assigned a unique number, and a random number generator (or random number table) selects n values. For example, to choose 10 students from a class of 30, number each student 01–30 and use a random number generator to pick 10 distinct numbers. The probability that any particular individual is selected is exactly n/N. SRS is the theoretical foundation for all other methods.

Systematic Sampling

Systematic sampling selects every k-th individual from a list, where k = N/n is called the sampling interval. A random starting point r is chosen from 1 to k, and then individuals at positions r, r+k, r+2k, … are selected. For example, if N = 500 and n = 50, then k = 10. If the random start is r = 4, you select individuals 4, 14, 24, 34, … This method is simple to implement but can introduce bias if the list has a periodic pattern that aligns with the interval k.

Stratified Sampling

When a population naturally divides into non-overlapping subgroups called strata (e.g., grade levels, age groups, genders), stratified sampling takes a separate random sample from each stratum. In proportional stratified sampling, the number of individuals chosen from each stratum is proportional to that stratum's share of the population. If stratum i contains N_i members and the total population is N, you sample n_i = n × (N_i/N) individuals from stratum i. This guarantees representation of every subgroup and typically produces more precise estimates than SRS of the same total size.

Cluster Sampling

Cluster sampling divides the population into groups called clusters (often defined by geography, like city blocks or schools), randomly selects a number of clusters, and then surveys all members within those chosen clusters. For example, to survey high schoolers across a large city, you might randomly pick 20 schools and survey every student in those schools. This is cost-effective when the population is spread over a wide area, but it can be less precise than SRS because individuals within a cluster tend to be more similar to each other than to the general population — a phenomenon measured by the intraclass correlation coefficient.

The Key Equation

P(individual selected)=nNP(\text{individual selected}) = \frac{n}{N}
  • P — Probability of selection (dimensionless): The likelihood, between 0 and 1, that a specific individual from the population ends up in the sample.
  • n — Sample size (count): The number of individuals chosen to be in the sample.
  • N — Population size (count): The total number of individuals in the population from which the sample is drawn.

Worked Example

Problem. A school has 800 students divided into four grade levels: Grade 9 (240 students), Grade 10 (200 students), Grade 11 (180 students), and Grade 12 (180 students). A researcher wants to take a proportional stratified random sample of 80 students. How many students should be sampled from each grade?

n9=24,  n10=20,  n11=18,  n12=18n_9=24,\; n_{10}=20,\; n_{11}=18,\; n_{12}=18

Answer. Sample 24 from Grade 9, 20 from Grade 10, 18 from Grade 11, and 18 from Grade 12.

Figure
Diagram
Random Sampling Methods ComparedPopulationSimple RandomAny 4 chosen equallySystematicEvery k-th elementStratifiedStratum AStratum BSample each groupClusterC1C2Select whole clustersFilled circles = selected individuals; shaded = not selected
Origins · 1 minJerzy Neyman and how this was discovered
JN
Jerzy Neyman
1894–1981 · Polish-American

Jerzy Neyman made foundational contributions to the theory of statistical sampling. In his landmark 1934 paper, he rigorously developed the theory of stratified sampling and introduced the concept of confidence intervals — proving mathematically why and when stratified sampling produces more precise estimates than simple random sampling. His framework gave researchers a principled way to choose between sampling strategies based on statistical efficiency. Neyman was originally trained as a mathematician in Poland and initially found statistics 'boring' — until he encountered real-world problems in agricultural field experiments that convinced him of its power and importance.

  • Sampling Bias — Understanding why random sampling matters requires knowing what goes wrong when samples are not random — sampling bias causes systematic errors that random methods are specifically designed to prevent.
  • Sample Size and Margin of Error — Once you have a random sample, sample size n directly determines the margin of error in your estimates, connecting sampling method to the precision of statistical conclusions.
  • Basic Probability — Random sampling is built on probability theory — the probability of selection formula P = n/N only makes sense once you understand what probability means and how to calculate it.
  • Observational Studies vs. Experiments — Random sampling is the key tool in observational studies; understanding how it differs from random assignment in controlled experiments clarifies when and why each design is appropriate.