Sample

A sample is a subset of individuals, observations, or data points selected from a larger population for the purpose of conducting a study. The sample provides the data from which researchers draw conclusions about the population, making sample selection one of the most consequential decisions in study design.

Because it is rarely feasible to study an entire population, researchers must rely on samples to estimate population parameters. The extent to which findings from a sample can be generalized to the broader population depends on how the sample was selected, its size, and its representativeness. A biased sample produces biased estimates that may not reflect the true state of the population, and sophisticated analysis can correct this only to the extent that the characteristics driving selection were measured. Understanding sampling methods and their limitations is essential for both designing studies and interpreting their results.

Sampling methods fall into two broad categories: probability sampling and nonprobability sampling. In probability sampling (including simple random, systematic, stratified, and cluster sampling), every member of the population has a known, nonzero chance of selection, enabling statistical inference from the sample to the population. In nonprobability sampling (including convenience, purposive, quota, and snowball sampling), selection is based on accessibility, judgment, or referral, and the resulting sample may not represent the population in quantifiable ways.

The sampling frame is the list (or map, register or other source) of members of the target population from which the sample is actually drawn; ideally it includes every member exactly once. An incomplete or inaccurate sampling frame introduces coverage error, excluding population segments that should have been eligible. The sampling method determines how individuals are selected from the frame.

Sample representativeness depends on both the sampling method and the response rate. Even a perfectly designed probability sample becomes biased if the response rate is low and nonrespondents differ systematically from respondents. Researchers assess representativeness by comparing sample characteristics (demographics, clinical features) to known population characteristics and by examining response rates.

In qualitative research, samples are typically small and purposively selected to maximize information richness and diversity of perspectives. Data saturation often guides sample size decisions in qualitative studies, although it is not the only rationale.

Use probability sampling when the goal is to produce estimates that generalize to a defined population with known precision. Use purposive or other nonprobability methods when the goal is to study specific subgroups, explore phenomena in depth, or when a sampling frame is unavailable. Always describe the sampling method, sampling frame, and achieved sample composition so readers can assess the representativeness and generalizability of the findings.

A common error is treating a convenience sample as if it were a probability sample, applying inferential statistics that assume random selection. Another mistake is failing to describe the sampling method and frame, leaving readers unable to assess representativeness. Researchers sometimes focus exclusively on sample size without considering sample composition, achieving a large but biased sample. Finally, ignoring nonresponse or dropout and analyzing only complete cases without assessing the potential for selection bias can produce misleading results.

A clinical epidemiology study aims to estimate the prevalence of chronic kidney disease in a rural population. The researchers use a two stage cluster sampling design, first randomly selecting 30 villages from a district roster and then randomly selecting 40 households per village. Within each household, one adult is randomly selected using a Kish grid, which in effect adds a third stage. Of 1,200 targeted adults, 1,068 complete the assessment (89% response rate). Comparison of the sample with census data shows close agreement on age, sex, and socioeconomic distribution, supporting the sample's representativeness on those characteristics. Because adults in larger households, and households in larger villages, had lower chances of selection, the analysis uses survey weights and standard errors that account for clustering by village.