Random Sampling & Assignment — I can explain random sampling and random assignment and their roles in valid conclusions. - Grade 11 Math

SECTION 1

Historical Context & Motivation

Long before modern statistics existed, people tried to draw conclusions about large groups by examining smaller portions. Ancient tax collectors in Egypt and Rome counted portions of harvests to estimate total yields, and census takers tried to tally entire populations—often with wildly inaccurate results. The fundamental challenge was always the same: how do you learn about a huge group without checking every single member? The answer, developed over centuries, turned out to hinge on a surprisingly simple idea—randomness.

1786 • Laplace's Population Estimate

Pierre-Simon Laplace used sample data from selected French communes to estimate the total population of France, pioneering the idea that a well-chosen subset could represent a whole country.

1908 • "Student's" t-Test

William Sealy Gosset, publishing under the pen name "Student," developed statistical methods for small samples while working at the Guinness brewery, proving that even small random samples yield reliable results when handled correctly.

1923 • Neyman's Sampling Theory

Jerzy Neyman published groundbreaking work formalizing random sampling, showing mathematically why every member of a population must have a known chance of being selected for survey results to be trustworthy.

1948 • First Randomized Clinical Trial

The British Medical Research Council conducted the first modern randomized controlled trial to test streptomycin for tuberculosis. Random assignment to treatment groups became the gold standard in medical research.

1996 • CONSORT Guidelines

The Consolidated Standards of Reporting Trials (CONSORT) statement was published, requiring researchers worldwide to clearly report how randomization was performed in experiments, cementing random assignment as essential to credible science.

Throughout this history, two distinct uses of randomness emerged. The first—random sampling—asks "Who should we study?" The second—random assignment—asks "How should we divide participants into groups?" Confusing these two ideas is one of the most common mistakes in statistics, and understanding the difference is the key to evaluating whether a study's conclusions are actually valid.

SECTION 2

Core Principles & Definitions

Before diving into diagrams and examples, let's nail down the foundational ideas. Both random sampling and random assignment use chance, but they serve completely different purposes in a study. Understanding each concept—and the type of conclusion it supports—will help you evaluate any research claim you encounter.

Random Sampling

A method of selecting participants from a population so that every individual has a known, nonzero probability of being chosen. This allows you to generalize findings from the sample to the entire population.

Random Assignment

A method of dividing participants who are already in a study into treatment groups using chance (like a coin flip). This allows you to establish cause-and-effect relationships.

Generalization

The ability to extend conclusions from a sample to the broader population it was drawn from. Generalization requires random sampling because non-random samples may be biased in ways we can't detect.

Causation

A conclusion that one variable directly causes a change in another. Causation requires random assignment because it ensures that the groups differ only in the treatment applied, eliminating confounding variables.

Confounding Variable

A hidden variable that influences both the explanatory and response variables, creating a false impression of a relationship. Random assignment distributes confounders evenly across groups so they cancel out.

✦

KEY TAKEAWAY

Think of it like cooking reviews on a food app. Random sampling is like reading reviews from a random mix of all customers—this tells you what the general public thinks (generalization). Random assignment is like randomly giving half the customers extra hot sauce and the other half none, then comparing their ratings—this tells you whether the hot sauce actually caused a change in satisfaction (causation). You need different tools for different questions.

SECTION 3

Visual Explanation — Sampling vs. Assignment

The diagram below illustrates the critical distinction between random sampling and random assignment. On the left side, you can see how random sampling draws a representative subset from a large population. On the right side, random assignment takes the participants already in a study and splits them into treatment groups using chance. Notice that random sampling addresses the question of who is in the study, while random assignment addresses what happens to them during the study.

Left: Random sampling selects individuals from a population so that the sample represents the whole group, enabling generalization. Right: Random assignment splits existing participants into treatment and control groups using chance, enabling causal conclusions.

Pay close attention to the arrows in the diagram. On the left, the arrow flows from the population down to the sample and then to generalization—random sampling determines the direction of that flow. On the right, the arrow splits the participants into two groups and then converges at a comparison, enabling a cause-and-effect conclusion. A study can use one, both, or neither of these techniques, and the type of conclusion the study can support depends entirely on which ones were used.

SECTION 4

How Randomness Creates Validity

You might wonder: why does randomness matter so much? The answer lies in the mathematical idea of expected value. When you randomly select from a population, the sample's average is an unbiased estimate of the population's average. Similarly, when you randomly assign participants to groups, the groups' characteristics are expected to be balanced—not because of any deliberate matching, but because randomness distributes differences evenly in the long run.

Why Random Sampling Prevents Selection Bias

Suppose you want to know the average amount of sleep high school juniors get. If you only survey students in your first-period class, those students might not be typical—maybe early risers self-select into first period, or maybe your school's schedule differs from others. This is called selection bias. Random sampling eliminates this problem because every student in the population has an equal chance of being selected. When every member has a known probability of inclusion, the mathematical framework of probability guarantees that the sample statistic converges toward the true population parameter.

UNBIASED ESTIMATOR

E(x̄) = μ

Where E(x̄) is the expected value of the sample mean, and μ is the true population mean. A random sample produces an unbiased estimate—on average, x̄ equals μ.

Why Random Assignment Eliminates Confounders

Now suppose you want to test whether a new study app improves test scores. If you let students choose whether to use the app, the students who choose it might already be more motivated—so you can't tell if any improvement was caused by the app or by pre-existing motivation. This hidden factor is a confounding variable. Random assignment solves this by distributing all variables—motivation, prior knowledge, sleep habits, everything—roughly equally between the treatment group and the control group. The only systematic difference between the groups is the treatment itself.

CAUSAL DIFFERENCE

Treatment Effect = x̄₁ − x̄₂

Where x̄₁ is the mean outcome of the treatment group and x̄₂ is the mean outcome of the control group. With random assignment, this difference is attributable to the treatment because confounders are balanced.

⚠

Common Misconception

Students often think that random sampling automatically means an experiment can prove causation. It doesn't! Random sampling tells you who to study (for generalization), while random assignment tells you how to structure the study (for causation). They are independent tools that answer different questions.

SECTION 5

The Four Study Scenarios

Every research study falls into one of four categories depending on whether it uses random sampling, random assignment, both, or neither. The type of valid conclusion you can draw depends entirely on which category the study falls into. The diagram below shows all four possibilities and the conclusions each supports.

The 2 × 2 grid of study designs. Only Scenario 2 (both random sampling and random assignment) supports both generalization and causal conclusions. Scenario 4 is the most common real-world experiment design—it can establish causation for those participants but cannot generalize to a broader population.

In the real world, Scenario 4 is extremely common. Most experiments use volunteers rather than randomly selected participants, so the researchers can establish cause-and-effect for the people in the study but must be cautious about generalizing to everyone. Meanwhile, Scenario 1 is typical of well-designed surveys and observational studies—they can tell us what a population looks like, but they can't prove that one variable causes another. The ideal Scenario 2 is rare because it's expensive and sometimes unethical to randomly select people and then randomly assign them to treatments.

SECTION 6

Worked Example — Evaluating a Study

Let's walk through a realistic scenario step by step to see how to identify the type of study and determine what conclusions are valid.

ℹ

Study Description

A school district wants to know if a new reading program improves literacy scores. They randomly select 200 students from all 3,000 eighth-graders in the district. They then randomly assign 100 of these students to the new reading program (treatment) and 100 to the existing program (control). After eight weeks, they compare the average reading scores.

Evaluating the Study Design

1
Step 1 — Identify the Population

The population is all 3,000 eighth-graders in the district. This is the group the researchers ultimately want to draw conclusions about.

Population = 3,000 eighth-graders in the district
2
Step 2 — Check for Random Sampling

The problem says the 200 students were randomly selected from all 3,000 eighth-graders. This means every student had a chance of being chosen, so random sampling was used. The results can be generalized to the entire population of eighth-graders in the district.

✓ Random sampling → Generalization is valid
3
Step 3 — Check for Random Assignment

The 200 selected students were then randomly assigned to either the new program or the existing program. This means any differences in prior ability, motivation, or other confounders are expected to balance out between the two groups.

✓ Random assignment → Causal conclusions are valid
4
Step 4 — Identify the Scenario

Since the study uses both random sampling and random assignment, it falls into Scenario 2—the gold standard. This is relatively rare in practice, which makes this a strong study design.

Scenario 2 — Both generalization and causation
5
Step 5 — State Valid Conclusions

If the treatment group scores significantly higher than the control group, the researchers can validly conclude that the new reading program caused the improvement (because of random assignment) and that this effect would likely hold for all eighth-graders in the district (because of random sampling).

"The new reading program causes improved literacy scores among eighth-graders in this district."

SECTION 7

Random Sampling vs. Random Assignment — Side by Side

Because these two concepts are so frequently confused, let's lay them out in a direct comparison. This table should serve as your quick-reference guide when evaluating any study.

Feature	Random Sampling	Random Assignment
What it controls	Who is selected to be in the study	Which group each participant is placed in
Purpose	Make the sample representative of the population	Make treatment groups equivalent to each other
Type of conclusion	Generalization	Causation
Eliminates	Selection bias	Confounding variables
Used in	Surveys, observational studies, polls	Experiments, clinical trials
Example method	Number every person in the population, use a random number generator to select participants	Flip a coin for each participant to decide treatment or control group
Without it, you cannot…	Extend results beyond the sample	Claim one variable caused a change in another

Key differences between random sampling and random assignment

✦

KEY TAKEAWAY

Here's a memory trick: Sampling = Scope and Assignment = Arrows. Random sampling broadens the scope of who your results apply to (the whole population). Random assignment lets you draw arrows of causation from one variable to another. Think: S for Sampling, S for Scope; A for Assignment, A for Arrows.

SECTION 8

Connections to Advanced Statistics

The concepts of random sampling and random assignment form the foundation for more advanced topics you'll encounter in AP Statistics and college-level courses. Understanding these basics now will make those future ideas much easier to grasp.

What You Know Now	What Comes Next
Random sampling makes the sample representative	Sampling distributions describe how sample statistics vary across all possible random samples, leading to confidence intervals and margin of error
Random assignment balances confounders	Hypothesis testing and p-values quantify whether an observed difference between groups is too large to be explained by the randomization alone
The four study scenarios determine valid conclusions	Meta-analysis combines results from many studies with different designs to build stronger overall evidence
Confounding variables threaten causal claims	Advanced methods like propensity score matching and regression discontinuity attempt to approximate random assignment when true experiments are impossible

How today's concepts connect to advanced statistical methods

In fields like medicine, psychology, and economics, researchers often face ethical or practical barriers to using both random sampling and random assignment. For instance, you can't randomly assign people to smoke or not smoke to study lung cancer. In these cases, statisticians have developed sophisticated quasi-experimental designs and observational study techniques that try to approximate the benefits of randomization. But the gold standard remains: random sampling for generalizability, random assignment for causality.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

Explain in your own words the difference between random sampling and random assignment. Why can't one substitute for the other?

Show answer

Random sampling is the process of selecting participants from a population so that every individual has a known chance of being included; it ensures the sample is representative and allows generalization. Random assignment is the process of dividing existing study participants into groups using chance; it ensures that confounders are balanced and allows causal conclusions. They cannot substitute for each other because they solve different problems: sampling addresses who is studied, while assignment addresses how the study is structured. A perfectly random sample that lacks random assignment can describe a population but cannot prove cause and effect.

Question 2: BASIC CALCULATION

A researcher has a list of 500 employees at a company and uses a random number generator to select 50 of them for a survey about workplace satisfaction. Is this random sampling, random assignment, or both? What type of conclusion can the researcher draw?

Show answer

This is random sampling only. Each of the 500 employees had a known chance of being selected (50/500 = 0.10 or 10%). Because there is no experiment—no treatment is being applied and participants aren't being divided into groups—there is no random assignment. The researcher can generalize the survey results to all 500 employees (e.g., 'approximately 72% of all employees are satisfied'), but cannot establish that any particular factor caused the level of satisfaction observed.

Question 3: INTERMEDIATE

A university researcher recruits 60 volunteers from an introductory psychology course. She randomly assigns 30 to listen to classical music while studying and 30 to study in silence, then compares their quiz scores. Identify (a) whether random sampling was used, (b) whether random assignment was used, (c) which scenario this study falls into, and (d) what valid conclusion can be drawn if the music group scores higher.

Show answer

(a) Random sampling was NOT used—the participants are volunteers from one specific course, not randomly selected from a larger population. (b) Random assignment WAS used—the 60 volunteers were randomly split into the music and silence groups. (c) This is Scenario 4 (random assignment only). (d) If the music group scores significantly higher, the researcher can conclude that classical music caused the improvement in quiz scores for this group of participants. However, she cannot generalize this conclusion to all college students or all people, because the participants were not randomly sampled from a broader population.

Question 4: APPLIED

A health department randomly selects 1,000 adults from the city's voter registration rolls and mails them a survey about exercise habits and heart health. The survey results show that people who exercise more than 3 hours per week have lower blood pressure. A local news station reports: 'Exercise causes lower blood pressure, study finds.' Is this headline valid? Explain your reasoning by identifying the study design and what conclusions it actually supports.

Show answer

The headline is NOT valid. The study used random sampling (1,000 adults randomly selected from voter rolls), which means the results can be generalized to all registered voters in the city. However, this is an observational study—no one was randomly assigned to exercise or not exercise. Without random assignment, there could be confounding variables: perhaps people who exercise more also eat healthier, sleep better, or have higher incomes that provide better healthcare. This is Scenario 1 (random sampling only). The valid conclusion is: 'Among registered voters in this city, people who exercise more tend to have lower blood pressure.' The word 'causes' should be replaced with 'is associated with' because only an association—not causation—can be established without random assignment.

Question 5: CRITICAL THINKING

Design a study to determine whether a new tutoring app improves math scores for all high school students in your state. Describe (a) how you would use random sampling, (b) how you would use random assignment, (c) what ethical or practical challenges you might face, and (d) why a study with both random sampling and random assignment is rarely conducted in education research.

Show answer

(a) To use random sampling, you would obtain a list of all high school students in the state and use a random number generator to select a representative subset—say 500 students. This ensures the results generalize to all high school students in the state. (b) To use random assignment, you would randomly assign half of the selected students (250) to use the tutoring app and the other half to continue with their regular study methods for a set period, then compare math scores. (c) Practical challenges include obtaining a complete list of all students, getting parental consent from randomly selected families (many may decline, introducing non-response bias), ensuring students in the control group don't access the app, and coordinating across many schools. Ethical concerns include potentially withholding a beneficial tool from control-group students. (d) This type of study is rare because random sampling across an entire state is logistically difficult and expensive, non-response rates can undermine the random sample, schools may refuse participation, and there are equity concerns about giving some students resources while denying others. Most education studies therefore use volunteers (Scenario 4), trading generalizability for feasibility.

SUMMARY

Lesson Summary

Summary

Random sampling and random assignment are two distinct uses of randomness in research, and each supports a different type of conclusion. Random sampling selects participants from a population so that every individual has a known chance of being chosen, enabling generalization from the sample to the whole population. Random assignment divides existing participants into groups using chance, balancing confounding variables and enabling cause-and-effect conclusions.

Every study falls into one of four scenarios: random sampling only (generalize but not causation), random assignment only (causation but not generalization), both (the gold standard), or neither (limited conclusions only). When you encounter any research claim, ask two questions: Was the sample randomly selected? Were participants randomly assigned to groups? The answers tell you exactly what conclusions are—and are not—supported by the data.