MCAT topics → Research Methods and Biostatistics → Sampling Methods (Random, Stratified, Cluster, Convenience)

Sampling Methods (Random, Stratified, Cluster, Convenience)

MCAT trap: Conflates simple random sampling with stratified random sampling. Stratified sampling divides the population into subgroups (strata) and randomly samples within each stratum to ensure representation of key subgroups.

Sampling methods are tested on the MCAT through passage-based study evaluation — and the most common confusion is treating convenience sampling as a threat to internal validity. It is not. Convenience sampling threatens external validity: the results may be real within the sample but can't be safely generalized to the broader population. Internal validity is about whether the study correctly establishes a causal relationship within its own design. The MCAT will describe a recruitment process — 'patients at an urban clinic,' 'undergraduate volunteers' — and expect you to name the method and identify the specific validity threat it creates.

The trickiest part is keeping stratified and cluster sampling distinct. Both involve dividing a population into groups, but their logic is opposite: stratified sampling divides to *guarantee* coverage of key subgroups, then randomly samples within each; cluster sampling divides into pre-existing groups and randomly selects *entire groups* for convenience. Students also consistently mix up which validity type convenience sampling threatens — it's external validity (can we generalize to the real population?), not internal validity (was the study itself well-controlled?).

The MCAT loves to present convenience sampling in medical research contexts — recruiting from a single clinic, using college students, or enrolling volunteers — and ask you to recognize that the findings may not generalize. This is a passage interpretation skill as much as a definitional one. When you see a study that enrolled 'patients presenting to an urban hospital' or 'volunteers who responded to a flyer,' that's your cue to flag external validity concerns.

Common misconceptions

Common mistake
Wrong: Stratified sampling is just another name for simple random sampling applied to the whole population.
Right: Stratified sampling divides the population into subgroups (strata) and randomly samples within each stratum to ensure representation of key subgroups.
Simple random sampling gives every individual in the population an equal chance of being selected — but by chance alone, key subgroups (e.g., women, elderly patients) might be underrepresented. Stratified sampling fixes this by first dividing the population into meaningful subgroups, then randomly sampling within each one, guaranteeing that each subgroup is represented. They both use randomization, but stratified sampling adds a deliberate structure to protect representation — they are not interchangeable.
Common mistake
Wrong: Convenience sampling primarily threatens internal validity by introducing confounding.
Right: Convenience sampling primarily threatens external validity (generalizability) by producing a non-representative sample, not internal validity.
Internal validity asks whether the study design correctly establishes a causal relationship — things like confounders and experimental control. Convenience sampling doesn't inherently introduce confounders within the study; it introduces a biased sample that may not reflect the broader population. That's an external validity problem: the results may be real within the sample but can't be safely generalized. Keep these straight: internal validity = did the study measure what it thinks it measured; external validity = do the findings apply outside the study.
Common mistake
Wrong: Cluster sampling and stratified sampling both ensure proportional representation of subgroups.
Right: Stratified sampling ensures subgroup representation by sampling within each stratum; cluster sampling randomly selects entire pre-existing groups and is used for logistical convenience, not guaranteed representation.
The key difference is intent. Stratified sampling is designed to *ensure* subgroup representation — you carve out strata and deliberately sample from each one. Cluster sampling randomly selects pre-existing groups (like schools, hospitals, or neighborhoods) and includes everyone in those groups — it's done for logistical efficiency, not to guarantee subgroup balance. A randomly selected cluster might over- or underrepresent certain demographics by chance, so cluster sampling doesn't protect representativeness the way stratified sampling does.
Free Deck audit

See if your Anki deck covers this topic.

Upload your deck →
Guided session

Stuck on this? An AI tutor that probes your understanding.

Start a session →

What the exam tests

  1. Know the definitions and trade-offs of all five methods: simple random, stratified, cluster, systematic, and convenience sampling — and be able to match a description in a passage to the correct method.
  2. Understand how each sampling method affects external validity and selection bias — specifically, recognize when convenience sampling makes a study's results ungeneralizable and why random sampling reduces selection bias.
  3. Read a passage description of how participants were recruited, identify the sampling method being used, and predict the specific bias or limitation that method introduces into the study's conclusions.

Can you avoid these mistakes?

A researcher studying hypertension recruits every third patient from a clinic's appointment list. What sampling method is this, and what validity concern does it raise?
A study on vaccine attitudes divides the U.S. population by age group (18–30, 31–50, 51+) and randomly selects 200 people from each group. How is this different from simple random sampling, and what problem does it solve?
A passage describes a study that enrolled 150 undergraduate psychology students as participants. What sampling method is this, and which type of validity — internal or external — is primarily threatened? Why?
A public health researcher randomly selects 10 counties from a state, then surveys all households in those counties. Is this stratified or cluster sampling? What is the researcher trading off by using this method?

Related topics

Study Designs (Cross-Sectional, Case-Control, Cohort, RCT) Variables (Independent, Dependent, Control, Confounding) Measures of Central Tendency (Mean, Median, Mode) Measures of Spread (Variance, SD, Range, IQR)

See how your Anki deck covers this topic.

Upload your deck for a free audit →