What is Sample Size? Definition & Statistical Power Principles
In empirical research and survey methodology, sample size represents the total count of individual participants, experimental subjects, or data observations collected to draw inferences about a broader population. Selecting an appropriate sample size is a vital step in experimental design: an undersized sample lacks the statistical power to detect real effects, while an oversized sample squanders financial resources and subjects participants to unnecessary testing.
The mathematical foundation of sample size determination derives from the Central Limit Theorem and William G. Cochran's pioneering 1953 sampling methodology. By balancing the desired margin of error, alpha significance level, statistical power, and population variability, researchers establish the precise sample size necessary to ensure statistical validity.
Statistical precision follows the law of diminishing returns: reducing the margin of error in half requires quadrupling the sample size. As sample size grows past approximately 1,000 respondents, additional data collection produces marginal gains in accuracy, explaining why nationwide political polls rarely survey more than 1,000 to 1,500 voters.
Interactive Probability Spectrum & Margin of Error Simulator
Dynamic distribution estimator calculating expected outcomes, standard error, and confidence interval bounds.
Mathematical Derivations and Calculation Formulas
Depending on population scale and whether the research targets a categorical proportion or continuous numeric mean, distinct algebraic formulas apply:
1. Cochran's Formula for Proportions (Infinite Population)
When estimating a population proportion with an infinite or unknown population size:
Worked Example: A market researcher targets a 95% confidence level (z = 1.960) with a ±4% margin of error (e = 0.04), assuming conservative maximum variance (p = 0.50):
2. Finite Population Correction (FPC)
When the target population N is small and known, unadjusted Cochran formulas overestimate sample requirements. Apply the Finite Population Correction:
Worked Example: Suppose the researcher above is surveying a private company with only N = 1,500 employees. Taking n = 601:
The FPC reduces the required sample from 601 down to 430, saving substantial time and administrative cost.
3. Sample Size for Continuous Means
When evaluating a continuous quantitative mean with anticipated population standard deviation σ and desired error tolerance E:
Statistical Power and Hypothesis Testing Considerations
In clinical trials and comparative scientific research, sample size must account for both false positive and false negative risk:
Sample Size Requirement Matrix (95% Confidence Level)
The reference matrix below highlights the interplay between population size N and target margin of error:
Core Parameters Influencing Sample Size
Every sample size calculation balances four foundational statistical levers:
Practical Real-World Applications
Sample size calculations govern empirical rigor across diverse fields:
- National Political Polling: Survey organizations survey roughly 1,000 to 1,500 voters to achieve a ±3% margin of error across the entire national electorate.
- A/B Testing & Conversion Optimization: Digital product teams compute required sample sizes before launching split tests to ensure statistical power without prolonging test duration.
- FDA & Clinical Trial Approvals: Pharmaceutical trials calculate statistical power to verify that therapeutic drug benefits can be distinguished from placebo effects.
- Industrial Quality Assurance: Manufacturers perform lot acceptance sampling to certify batches against acceptable quality levels (AQL), paired with our Confidence Interval Calculator and Probability Calculator.
Cochran's Sampling Theorem, Power Analysis & Finite Population Multipliers
Determining optimal scientific sample size is governed by William G. Cochran's sampling formula, balancing experimental statistical power (1 − β) against type I false positive error rates (α). Setting the population proportion p = 0.5 represents the maximally conservative minimax strategy, maximizing variance p(1 − p) = 0.25 and guaranteeing sufficient sample representation regardless of true underlying demographic splits. When the target population N is finite and small relative to sample size n (typically n/N > 0.05), applying the Finite Population Correction (FPC) factor √((N − n) / (N − 1)) substantially reduces required sample size without sacrificing statistical confidence or margin of error bounds.
Frequently asked questions
What is sample size in statistical research?
Sample size (denoted by n) is the total number of individual observations, participants, or experimental units sampled from a broader population. Determining the optimal sample size ensures that statistical findings are mathematically valid, reliable, and sufficiently powered without wasting excess resources.
How do you calculate sample size for an infinite or unknown population?
Use Cochran's formula: n = [ z² × p(1 − p) ] ÷ e², where z is the critical value for the chosen confidence level, p is the estimated population proportion (use 0.5 for maximum conservative sample size), and e is the desired margin of error (expressed as a decimal).
What is the Finite Population Correction (FPC)?
When the population N is small and known (specifically when sample size n exceeds 5% of population N), the Finite Population Correction adjusts the required sample size downward: n_adj = n ÷ [ 1 + ((n − 1) ÷ N) ]. This accounts for the fact that sampling a large fraction of a finite group reduces sampling variance.
Why is p = 0.5 (50%) used as the default proportion in sample size formulas?
The product p(1 − p) reaches its mathematical maximum at p = 0.5 (0.5 × 0.5 = 0.25). Assuming p = 0.5 yields the most conservative (largest) possible sample size, guaranteeing sufficient statistical power regardless of the true underlying proportion.
What is statistical power (1 − β)?
Statistical power is the probability of correctly rejecting a false null hypothesis (detecting a true real-world effect when one actually exists). Standard convention targets a minimum statistical power of 0.80 (80%) or 0.90 (90%) in clinical trials and scientific studies.
How does margin of error impact sample size?
Sample size is inversely proportional to the square of the margin of error (e²). Halving the margin of error (e.g., from ±6% down to ±3%) requires quadrupling the sample size (2² = 4 times as many participants).
Why do national political polls typically survey around 1,000 respondents?
Under Cochran's formula, a sample size of n ≈ 1,067 yields a ±3% margin of error at a 95% confidence level for an infinite population. Beyond 1,000 respondents, diminishing returns set in: reducing the margin of error to ±1% would require over 9,600 respondents.
How does population size N affect required sample size?
For large populations (e.g., N > 50,000), total population size has virtually zero impact on the required sample size. A sample of 1,000 provides the same ±3.1% margin of error whether sampling a city of 100,000 or a country of 330 million.