Home Statistical Dictionary Sample Size Determination
Sample Size Determination
The broader process of choosing how many units to sample or enroll in a study, incorporating not only statistical power requirements but also precision targets for confidence intervals, expected nonresponse or attrition, design effects, and practical budget constraints.
In Plain English
Sample size determination is the full, practical exercise a study designer goes through to land on a final target number of participants or units to sample, and it's broader than power analysis alone. Power analysis answers 'how big a sample do I need to reliably detect this effect in a hypothesis test?', but sample size determination folds in everything else that affects the number you actually need to aim for in the field: how much a complex survey design (clustering, stratification) will inflate your effective sample size requirement via the design effect, how much of your sample you expect to lose to nonresponse or attrition before you even get to analyze it, whether you're planning to estimate a confidence interval with a specific target width rather than run a formal hypothesis test at all, and, very concretely, how much money and time you actually have to work with.
Definition
Sample size determination is the overall process of specifying the target number of units to sample or enroll in a study, incorporating power analysis (the sample size needed to detect a specified effect at a target power level) as one input among several others: precision-based targets (the sample size needed for a confidence interval of a specified maximum width, common in survey estimation where no formal hypothesis test is planned), the design effect (inflating the required sample size to compensate for clustering or stratification relative to an equivalent simple random sample), anticipated nonresponse or attrition rates (inflating the initial sample size drawn so that the final achieved, analyzable sample still meets the target after expected losses), and practical constraints of budget, time, and logistics that may ultimately cap the feasible sample size below what a pure statistical calculation would recommend. Sample size determination is thus the umbrella planning process within which power analysis operates as one, often central, but not exclusive, statistical component, the final target sample size in a real study protocol typically reflects the interaction of all these factors together, not a power calculation performed in isolation.
Properties
- A common practical formula for sample size determination inflates a statistically derived target sample size by two multiplicative factors before it becomes the actual field target, the design effect (to compensate for the precision loss from clustering or other design complexity) and the inverse of the expected response or retention rate (to compensate for anticipated nonresponse or attrition), producing a substantially larger initial sample than the underlying power or precision calculation alone would suggest.
- Survey estimation problems that are purely descriptive, estimating a population proportion or mean within a target margin of error, rather than testing a formal hypothesis, use precision-based sample size formulas rather than power-analysis formulas, both approaches share the same underlying statistical logic but are framed around different planning goals, a target confidence interval width versus a target probability of detecting a specified effect.
- Real-world sample size determination is frequently constrained as much by budget, timeline, and logistical feasibility as by pure statistical calculation, a well-documented tension in applied research, where the statistically ideal sample size and the practically achievable sample size often diverge, requiring researchers to either secure additional resources or explicitly report the resulting limitations on the study's statistical power or precision.
At a Glance