Sampling methods

Sampling Fundamentals - The Right Slice

• Population (N): The entire group a study aims to understand.
• Sample (n): A representative subset of the population from which data is collected.
  • The goal is to make inferences about the population.
• Sampling Frame: The specific list of individuals from which the sample is drawn (e.g., a clinic's patient list).
• Sampling Bias: A systematic error where the sample is not representative of the population, threatening the study's external validity.

⭐ Generalizability (External Validity): The degree to which findings can be applied to the broader population. This is highly dependent on how well the sample represents the population.

Probability Sampling - Truly Random Acts

Ensures every member of the population has a known, non-zero chance of being selected, minimizing selection bias. Essential for generalizability (external validity).

• Simple Random Sampling (SRS)

  • Every individual has an equal chance of selection.
  • Like a lottery; requires a full population list (sampling frame).

• Systematic Sampling

  • Select individuals at a regular interval (every k-th person) from a list after a random start.
  • Efficient, but can be biased if the list has a periodic pattern.

• Stratified Sampling

  • Divide population into homogeneous subgroups (strata), e.g., by age or race.
  • Perform SRS within each stratum.
  • Guarantees representation of key subgroups.

• Cluster Sampling

  • Divide population into heterogeneous groups (clusters), e.g., hospitals or zip codes.
  • Randomly select entire clusters to sample.
  • 📌 Mnemonic: "Clusters are mini-populations."

⭐ Stratified sampling increases precision and ensures minority subgroups are adequately represented, boosting statistical power for subgroup analyses.

Non-Probability Sampling - Conveniently Biased

Selection isn't random; it relies on the researcher's judgment or convenience. This introduces selection bias, limiting the generalizability of findings to the broader population.

• Types of Non-Probability Sampling:
  • Convenience Sampling: Choosing easily accessible subjects (e.g., patients in a single clinic). Very prone to selection bias.
  • Quota Sampling: Filling pre-set quotas for subgroups (e.g., 50 men, 50 women) in a non-random way.
  • Purposive (Judgmental) Sampling: Researcher handpicks subjects based on specific criteria or expertise.
  • Snowball Sampling: Participants recruit other eligible participants. Useful for hard-to-reach or hidden populations.

⭐ Key limitation: Because the sample is not representative, findings from non-probability sampling cannot be generalized to the entire population. The study has low external validity.

Sampling Biases - Dodging Disasters

• Selection Bias: Sample is not representative of the target population, limiting external validity.
  • Ascertainment Bias: Nonrandom sampling creates a skewed sample (e.g., using only hospitalized patients).
  • Nonresponse Bias: Participants differ significantly from non-participants.
  • Berkson Bias: Hospital-based samples show higher disease prevalence vs. general population.
  • Healthy Worker Effect: Working populations are healthier than the general population.

⭐ Neyman (Prevalence-Incidence) Bias: In case-control studies, missing severe or rapidly fatal cases leads to a non-representative sample.

High‑Yield Points - ⚡ Biggest Takeaways

• Random sampling is crucial for generalizability (external validity), allowing inferences about a larger population.
• Stratified sampling ensures specific subgroups are adequately represented, improving precision for those groups.
• Cluster sampling randomly selects natural groups (e.g., hospital wards), offering convenience but with lower precision.
• Convenience sampling is highly susceptible to selection bias, severely limiting external validity.
• Random sampling minimizes selection bias; randomization in trials minimizes confounding.