Sampling techniques

Sampling Fundamentals - The Pick of the Litter

• Simple Random: Every individual has an equal chance of selection. Highly representative but can be impractical.
• Systematic: Select subjects at a regular interval (every kth person). Simple but risks periodicity bias.
• Stratified: Divide population into subgroups (strata), then randomly sample from each. Guarantees minority representation.
• Cluster: Randomly select natural groups (e.g., hospital wards), then sample all individuals within.

⭐ Cluster sampling is cost-effective but increases standard error if intra-cluster correlation is high.

Probability Sampling - Every Ticket Wins

In probability sampling, every member of the population has a known, non-zero chance of being selected. This method is crucial for minimizing selection bias and allowing for statistical inference to the broader population.

• Simple Random Sampling
  • Each individual has an equal probability of selection.
  • Analogous to a lottery draw; ideal for homogenous populations.
• Systematic Sampling
  • Selection of every n-th person from a list (e.g., every 5th patient).
  • ⚠️ Risk of bias if the list has an underlying periodic pattern.
• Stratified Sampling
  • Population is divided into homogenous subgroups (strata), e.g., by age or disease severity.
  • Simple random or systematic sampling is then performed within each stratum.
  • Ensures representation of key subgroups.
• Cluster Sampling
  • Population is divided into clusters (e.g., hospitals, cities).
  • Entire clusters are randomly selected, and all individuals within those clusters are studied.

⭐ Stratified sampling increases precision and is superior to simple random sampling when you need to ensure that minority subgroups are adequately represented in the sample.

Non-Probability Sampling - The Usual Suspects

• Selection is non-random, based on convenience or specific criteria, not chance.
• High risk of selection bias, limiting generalizability to the entire population.

• Convenience Sampling: Selecting whoever is easiest to reach (e.g., patients in your clinic).
• Purposive (Judgmental) Sampling: Researcher hand-picks subjects with specific expertise or characteristics.
• Snowball Sampling: Existing subjects recruit future subjects from among their acquaintances. Useful for hidden populations (e.g., IV drug users).
• Quota Sampling: A non-random method to ensure certain subgroups are represented in the sample.

⭐ High-Yield: The inability to calculate a sampling error is a key drawback. You cannot be confident that your sample's findings reflect the true population values.

Sampling Pitfalls - Dodging Disasters

• Selection Bias: Sample is not representative of the target population, limiting generalizability.
  • Berkson Bias: Hospital-based samples show ↑ disease prevalence vs. community.
  • Neyman Bias (Prevalence-Incidence): Excludes severe/fatal cases, underestimating risk.
  • Volunteer Bias: Participants are often healthier or more motivated.
• Non-response Bias: Significant differences between those who participate and those who do not, skewing results.
• Recall Bias: Inaccurate recollection of past exposures; a major issue in case-control studies.
• Observer Bias (Hawthorne Effect): Subjects change behavior because they are being observed.

⭐ The Hawthorne effect can cause an overestimation of an intervention's true efficacy, as participants may improve simply from the attention of being in a study.

High-Yield Points - ⚡ Biggest Takeaways

• Simple random sampling gives every individual an equal chance of selection, maximizing generalizability.
• Stratified sampling divides the population into subgroups (strata) and samples from each to ensure minority representation.
• Cluster sampling randomly selects natural groups (clusters) and samples all members; it is efficient but has a higher sampling error.
• Systematic sampling selects subjects at a regular interval, risking bias from underlying list patterns.
• Non-random sampling is prone to selection bias, severely limiting external validity.